MicrostockGroup
Agency Based Discussion => Adobe Stock => Topic started by: MatHayward on December 05, 2022, 09:25
-
Hi everyone,
Today, we are announcing our policy regarding generative AI content, and I’d like to share that we have begun accepting illustrations made using generative AI into our collection. We believe that generative AI tools can help our contributor community continue to create amazing content, and we believe in transparent, clear labeling for customers when it comes to this content.
We have prepared generative AI content submission guidelines and a page to answer common questions. We believe that our policy to accept AI generated content will enable contributors and customers to benefit from the value that AI generated content can bring.
In our Discord channel for Adobe Stock contributors, we opened a new channel #ai-generated-talk as a forum for addressing further questions. As always, I will also monitor this thread daily and will do my best to answer any questions not covered in the FAQ.
Submission guidelines: https://helpx.adobe.com/stock/contributor/help/generative-ai-content.html
FAQ: https://helpx.adobe.com/stock/contributor/help/generative-ai-faq.html
Discord community: https://discord.com/invite/adobestock
Thank you,
Mat Hayward
-
Hi again, moments ago, a blog article was published with more detailed information on this important announcement. You can read it here: https://blog.adobe.com/en/publish/2022/12/05/amplifying-human-creativity-adobe-stock-defines-new-guidelines-content-generative-ai
Let me know if you have any questions.
-Mat Hayward
-
Hi again, moments ago, a blog article was published with more detailed information on this important announcement. You can read it here: https://blog.adobe.com/en/publish/2022/12/05/amplifying-human-creativity-adobe-stock-welcomes-content-using-generative-ai
Let me know if you have any questions.
-Mat Hayward
Mat, link doesn't seem to work?
-
Mat, link doesn't seem to work?
Try this: https://blog.adobe.com/en/publish/2022/12/05/amplifying-human-creativity-adobe-stock-defines-new-guidelines-content-generative-ai
-
Hi again, moments ago, a blog article was published with more detailed information on this important announcement. You can read it here: https://blog.adobe.com/en/publish/2022/12/05/amplifying-human-creativity-adobe-stock-defines-new-guidelines-content-generative-ai
Let me know if you have any questions.
-Mat Hayward
I am not sure but should we be happy or sad?
With AI content being accepted, it has opened doors for maximum non-qualified people to upload the assets.
Where are we artist now?
-
Hi again, moments ago, a blog article was published with more detailed information on this important announcement. You can read it here: https://blog.adobe.com/en/publish/2022/12/05/amplifying-human-creativity-adobe-stock-defines-new-guidelines-content-generative-ai
Let me know if you have any questions.
-Mat Hayward
I am not sure but should we be happy or sad?
With AI content being accepted, it has opened doors for maximum non-qualified people to upload the assets.
Where are we artist now?
The whole point is to get rid of artists, a major expense for AdobeStock. What I don't understand is who they think will be subscribing to their software when we are all out of business. Ironic that it's Getty and SS that seem to be at least a bit concerned with the ramifications.
-
Hi again, moments ago, a blog article was published with more detailed information on this important announcement. You can read it here: https://blog.adobe.com/en/publish/2022/12/05/amplifying-human-creativity-adobe-stock-defines-new-guidelines-content-generative-ai
Let me know if you have any questions.
-Mat Hayward
Hi Mat,
thank you very much! It seems a quite serious approach and a way to clarify all the doubts expressed here.
I think too that this is a new creative tool in our hands.
One questions please:
if the AI generated image is used as a sketch and heavily processed after, must we always write made with generative AI?
Thanks in advance!
-
Hi again, moments ago, a blog article was published with more detailed information on this important announcement. You can read it here: https://blog.adobe.com/en/publish/2022/12/05/amplifying-human-creativity-adobe-stock-defines-new-guidelines-content-generative-ai
Please refer to the FAQ for additional information: https://helpx.adobe.com/stock/contributor/help/generative-ai-faq.html
-Mat Hayward
Hi Mat,
thank you very much! It seems a quite serious approach and a way to clarify all the doubts expressed here.
I think too that this is a new creative tool in our hands.
One questions please:
if the AI generated image is used as a sketch and heavily processed after, must we always write made with generative AI?
Thanks in advance!
Thanks for the question, @gameover. Yes, even when your images are heavily processed in post, it is important you tag the content with "generative AI" in the title and in the keywords (in English). In addition, you must include "generative" and "AI" as two separate keywords.
Thanks again,
Mat Hayward
-
Where are we artist now?
Stock agencies were never a great place for artists anyway. Now even more so.
-
Hi Matt, thanks for the information.
One question: Works that contain a combination of AI and 3d rendering should also be titled and tagged as AI?
-
Hi Matt, thanks for the information.
One question: Works that contain a combination of AI and 3d rendering should also be titled and tagged as AI?
Please refer to the FAQ for detailed information on metadata requirements: https://helpx.adobe.com/stock/contributor/help/generative-ai-faq.html
thanks,
Mat Hayward
-
OK thank you
-
I am not sure but should we be happy or sad?
With AI content being accepted, it has opened doors for maximum non-qualified people to upload the assets.
Where are we artist now?
if their work is accepted, how are they non-qualified? what makes you an 'artist' and not them?
early photograpy was rejected by 'real artists'
it's evolution in action
-
The whole point is to get rid of artists, a major expense for AdobeStock. What I don't understand is who they think will be subscribing to their software when we are all out of business. Ironic that it's Getty and SS that seem to be at least a bit concerned with the ramifications.
i havent seen them considering anything other than the copyright questions on the training sets -- SS is accepting AI art made with their database
-
And does anyone even know what programs can generate these images? And what are the PC requirements?
-
there're some confusing stds listed:
2 very different requirements for tagging:
https://helpx.adobe.com/stock/contributor/help/generative-ai-content.html
Do: Specify that these depictions are fictitious and generated
https://helpx.adobe.com/stock/contributor/help/generative-ai-faq.html
Tag your generative AI content to help customers find it:
• Titles should include the phrase, in English, "Generative AI"
- Example title: Beautiful landscape. Generative AI.
• Keywords should include the English phrase "Generative AI" as well as "Generative" and "AI" in the language you are submitting your assets.
- Example keywords: Landscape, No People, Generative AI, Generative, AI.
so tags have to include all these:
fictitious, generated ,"Generative AI" , "Generative" , "AI" ??
-----------------------------
"Don’t tag generative AI content with real place names."
what is considered specific? castle in Wales, Everest, Taj Mahal, Matterhorn??
---------------
I submitted "Imaginary ancient Egyptian papyrus of Horus from the Book of the Dead - original CGI illustration "
rejected saying it needed a property release - i questioned & was told to re-submit, but concerned next reviewer would still reject
-
I am not sure but should we be happy or sad?
With AI content being accepted, it has opened doors for maximum non-qualified people to upload the assets.
Where are we artist now?
if their work is accepted, how are they non-qualified? what makes you an 'artist' and not them?
early photograpy was rejected by 'real artists'
it's evolution in action
I think you are missing the point. There is difference between create and generate.
-
And does anyone even know what programs can generate these images? And what are the PC requirements?
We've been through this.
-
When the Daguerreotype appeared and photography was born as an art form many painters turned against photographers saying that there was no such thing as an art form in photos because photographers were just copying reality. It is curious that Reality was what most painters were painting at that time.
Then with digital and Sw like photoshop, many followed the wave of saying that those images were not art photography. They were pure manipulation of reality, forgetting also the principle of image value (representation of an idea or a presented idea). AI technology allows to improve an artist's performance in many ways and brings new possibilities for an artist to express himself. It is in the Prompt that we identify the originality and value of the represented image.
Therefore a person that understands code and use different sw to produce images is just valid to me as a painter.
There is a fine essay that is taught in Art Universities called "The Work of Art in the Age of Mechanical Reproduction" from Walter Benjamin which helps to understand that Generative process is just that: a process. Oil painting, analogue or digital photography and even generative AI are not exactly copy/paste but an process of making images.
There are lots of SW online to make "G-Art". We all can go to Github and craft our own SW based in the available codes that are free there. Also we can use sw like stable diffusion to mix, create, train our own models like most of people do nowadays.
Nevertheless, in short: I do like and make "G-Art". Some things are amazing but still... i prefer doing street photography. No code can give me the pleasure to shoot those moments with my camera.
-
Let's take a real-life situation (at least it's real-life for me) : suppose I "create" a beautiful background of romantic woodlands, or a victorian interior, or anything in Midjourney (or any other AI program) and I use it as a background for a cut-out photo of a real living model. So the model is a photo and needs a model release, and the background is AI generated.
If I put this image in my Adobe portfolio for sale, it should be submitted as an illustration with AI Generated in title and keywords. And because it is AI generated, reviewers will assume the woman/model is also AI generated, so not needing a model release ...
I think I already know the answer : submit as illustration and add a MR, but how will this affect search? If a buyer searches for a model in romantic woodlands, and he/she ticks the "with people" button, which images will appear? Only the ones with MR, or also the images with an AI person in it ? Interesting problem?
Would like to hear Matt's opinion about this ...
-
When the Daguerreotype appeared and photography was born as an art form many painters turned against photographers saying that there was no such thing as an art form in photos because photographers were just copying reality. It is curious that Reality was what most painters were painting at that time.
Those are very good points to be honest. The artists who're going to hurt the most are the ones (and have always been the ones) who have devoted many years to learning and perfecting a craft only to find it obsolete one day and has been taken over by a completely different process (painting to photography, analog to digital etc.). So I can understand why people feel hurt at technological shifts in artistic process because when you've invested so much in one particular process it's too difficult to change course and learn another. Especially when your livelihood depends on it. I mean, I can understand why many people would be intimidated by having to learn to code to make art in the future because it is not an easy thing to learn at all.
-
Hi everyone,
Today, we are announcing our policy regarding generative AI content, and I’d like to share that we have begun accepting illustrations made using generative AI into our collection.
Thanks Mat, that's the right way to manage the new AI technology as a fantastic tool to produce great images!
Other agencies come out with non sense rules like "you can't produce with AI, but we can" while Adobe make the right choice for contributors.
AI will of course deeply change the photography history, but I think it's unavoidable process, impossible to stop.
Next years will be challenging, and that's the life. Good luck everyone
-
Hi everyone,
Today, we are announcing our policy regarding generative AI content, and I’d like to share that we have begun accepting illustrations made using generative AI into our collection.
Thanks Mat, that's the right way to manage the new AI technology as a fantastic tool to produce great images!
Other agencies come out with non sense rules like "you can't produce with AI, but we can" while Adobe make the right choice for contributors.
AI will of course deeply change the photography history, but I think it's unavoidable process, impossible to stop.
Next years will be challenging, and that's the life. Good luck everyone
I think will be like HDR. First year extra popular then every year less. We will see..
-
The whole point is to get rid of artists, a major expense for AdobeStock. What I don't understand is who they think will be subscribing to their software when we are all out of business. Ironic that it's Getty and SS that seem to be at least a bit concerned with the ramifications.
i havent seen them considering anything other than the copyright questions on the training sets -- SS is accepting AI art made with their database
I agree, hence "at least a bit concerned".
I suspect Adobe is ignoring copyright implications as they will eventually just buy an AI solution or make their own. They don't want to set a precedent for compensating artists.
I think they want to be a software provider making their money from subs to their AI engine (or their partners). This will make up for less demand for their traditional offering.
Getty and SS may end up licensing their collections to other companies for training (as well as saving on payouts to their own contributors).
Just spitballing.
-
Very sensible policy! Thank you!
-
I imagine that with images produced by AI the inventory of images would overflow, even more. The higher the supply, the lower the prices. Well, for those who are lucky to sell! Very excited about AI image generation programs, welcome to the progress! I love it! :)
-
What is best soft for use? It was any limitation with software brand?
-
Did a search on illustrations at various agencies. If you sort by new, it looks like a very large percentage is already being created with ai.
So it is good to have a clear path forward.
It is still work, I need many different prompts and then photoshop to get the results I want, even on very, very simple images.
I don‘t think ai will replace us, because I don‘t think customers can quickly generate the images they need if they have no creative background.
But for creatives this is an opportunity. If you know how to draw, but don‘t know how to do an oil painting, the ai can help you with that.
Great at watercolors, but cannot do charcoal? The ai can help you with that.
But unless you have professional creative training or background, i think it will be extremely difficult to create sellable content.
But for the creatives at Adobe this is a great opportunity to do something new and have fun experimenting.
I have been playing around with various ai‘s for two years now, would be great if I could earn back all the credits I spent.
-
I think this is all actually really funny. Adobe is scared of missing the boat and is now accepting AI images.
But, first of all they are trying to put the risk at you. They state that artists supplying the images should have all the rights to the picture so that in the end they could not be held responsible if there were any copyright issues and stuff. But that won't work obviously. It's Adobe that sells the picture to the client, not the artist. Any claim will come to them first. Sure they can try to get damages back from the artist but most of the time they will not have the funds to cover these damages.
Secondly, this whole AI stuff will not be moved anymore via microstock agencies. Clients will license the appropriate tool and produce the images on their own. There will be no need for agencies anymore. So, is this Adobe's last attempt grabbing the money (as others) before their role is played out if AI really takes off?
What is your take on that Mat?
-
there're some confusing stds listed:
2 very different requirements for tagging:
https://helpx.adobe.com/stock/contributor/help/generative-ai-content.html
Do: Specify that these depictions are fictitious and generated
https://helpx.adobe.com/stock/contributor/help/generative-ai-faq.html
Tag your generative AI content to help customers find it:
• Titles should include the phrase, in English, "Generative AI"
- Example title: Beautiful landscape. Generative AI.
• Keywords should include the English phrase "Generative AI" as well as "Generative" and "AI" in the language you are submitting your assets.
- Example keywords: Landscape, No People, Generative AI, Generative, AI.
so tags have to include all these:
fictitious, generated ,"Generative AI" , "Generative" , "AI" ??
-----------------------------
"Don’t tag generative AI content with real place names."
what is considered specific? castle in Wales, Everest, Taj Mahal, Matterhorn??
---------------
I submitted "Imaginary ancient Egyptian papyrus of Horus from the Book of the Dead - original CGI illustration "
rejected saying it needed a property release - i questioned & was told to re-submit, but concerned next reviewer would still reject
Well there go five tags from the 49?
If I upload a photo I took of (place name here) which I filtered, and then create a version, using AI, then I can't tag it as (place name here) because it's a "fictitious", "generated", "Generative AI" , "Generative" , "AI" Image? And we can't use a real place name to help ID the original subject for someone? Or maybe we should be required to name the place because it's based on a real subject? That's confusing.
I'm not going to touch that Public Domain or out of copyright subject, because Adobe doesn't take those? :) Honestly, there does seem to be a time period where antique, becomes ancient and they do, but unlike IS that says 1900, AS just says they are not allowed.
I like that they have come out with some guidelines and ideas how we might be able to contribute AI assisted images.
This is going to end up being a case for the lawyers and courts, and maybe Adobe doing this will press the issue instead of running away scared, and we'll get some clarity to what's allowed and what's not. I mean by the law, not from some agency lawyer who decides what his opinion is. We need legal precedent to start making the limits and rights of our images, clear.
-
I think this is all actually really funny. Adobe is scared of missing the boat and is now accepting AI images.
But, first of all they are trying to put the risk at you. They state that artists supplying the images should have all the rights to the picture so that in the end they could not be held responsible if there were any copyright issues and stuff. But that won't work obviously. It's Adobe that sells the picture to the client, not the artist. Any claim will come to them first. Sure they can try to get damages back from the artist but most of the time they will not have the funds to cover these damages.
Secondly, this whole AI stuff will not be moved anymore via microstock agencies. Clients will license the appropriate tool and produce the images on their own. There will be no need for agencies anymore. So, is this Adobe's last attempt grabbing the money (as others) before their role is played out if AI really takes off?
What is your take on that Mat?
Yes, it is not clear that with copyrights. There may be any problems later, and stock agencies may ban them later. The problems are:
1. The developer of the AI program will say that all the images created by this program are his authorship.
2. The AI program will create something based on the existing one, and the author of the image will sue you.
I also do not understand how you can generate something and say that this is your job. We need a legal basis, what the law says about such images, who they belong to, who their author is. I'm afraid to create such images and upload them to adobe.
Adobe should give more legal advice on these matters.
-
Or maybe, it's just that Adobe cannot distinguish an AI image from a real one and that's why they are doing this?
But then they have admitted accepting them, which does not make their case in court any stronger if the sh*t hits the fan.
Even though they have a way to sue the artist who submitted the image to begin with since you promised you have all the rights :)
-
Or maybe, it's just that Adobe cannot distinguish an AI image from a real one and that's why they are doing this?
Mmmm... Adobe use its own AI in its software, it's in Lightroom and Photoshop, so I think their legal dept knows well the issue about it. They know much more than shutterstock or others. It would be really strange to think that they don't know and cannot disinguish an AI image from others.
By the way, it's curious that no one ask how Adobe trained its own AI that they put and offer to use in its own applications ;-)
Adobe is scared of missing the boat and is now accepting AI images.
Scared for missing the boat? Well, if so, they could do as others, telling us that submit AI images is against their rules.
I think the opposite: Adobe is one step ahed other agencies, they know what is AI and the legal implications, so they decide to cross the river and give contributors ability to use the technology.
Personally I like this way, much better than tell us "you can't use it, only us can use the new technology"
-
I want to be able to opt out my images for use in AI training.
-
Hi everyone,
Today, we are announcing our policy regarding generative AI content, and I’d like to share that we have begun accepting illustrations made using generative AI into our collection. We believe that generative AI tools can help our contributor community continue to create amazing content, and we believe in transparent, clear labeling for customers when it comes to this content.
We have prepared generative AI content submission guidelines and a page to answer common questions. We believe that our policy to accept AI generated content will enable contributors and customers to benefit from the value that AI generated content can bring.
In our Discord channel for Adobe Stock contributors, we opened a new channel #ai-generated-talk as a forum for addressing further questions. As always, I will also monitor this thread daily and will do my best to answer any questions not covered in the FAQ.
Submission guidelines: https://helpx.adobe.com/stock/contributor/help/generative-ai-content.html
FAQ: https://helpx.adobe.com/stock/contributor/help/generative-ai-faq.html
Discord community: https://discord.com/invite/adobestock
Thank you,
Mat Hayward
Well done Adobe! I am proud of your response to the new emerging technology. It is after all, another tool to be used by creatives. Thank you!
-
Seems like a much better response than others like SS which have stopped people from uploading AI art, and then have turned around and said that they will make it themselves, giving a very unclear payment structure method.
I guess its evolution, and its upto the artist to learn how to best manage this tool and get something amazing. Its the same as saying that just buying the most expensive DSLR does not automatically make you an artist same will be with this.
The better question is where do we start learning? And what software to pick up?
-
But, first of all they are trying to put the risk at you. They state that artists supplying the images should have all the rights to the picture so that in the end they could not be held responsible if there were any copyright issues and stuff.
To be fair, they've been doing that with images, videos and vectors since the beginning.
-
@MatHayward
May I check a couple of points please?
I read in the instructions that keywords for a location shouldn't be included. What should I do if I am basing the generated AI on my own uploaded photo of a specific location. If the image is clearly of the Pyramids in Egypt or the Eiffel Tower, as an example, can these keywords be included?
If the AI software is used on my own image purely as a means of transforming it into a digital painting, does this still get submitted as AI content, or can it be submitted just a an illustration?
Thanks
-
But, first of all they are trying to put the risk at you. They state that artists supplying the images should have all the rights to the picture so that in the end they could not be held responsible if there were any copyright issues and stuff.
To be fair, they've been doing that with images, videos and vectors since the beginning.
True, but it seems very explicit now.
-
Or maybe, it's just that Adobe cannot distinguish an AI image from a real one and that's why they are doing this?
Mmmm... Adobe use its own AI in its software, it's in Lightroom and Photoshop, so I think their legal dept knows well the issue about it. They know much more than shutterstock or others. It would be really strange to think that they don't know and cannot disinguish an AI image from others.
By the way, it's curious that no one ask how Adobe trained its own AI that they put and offer to use in its own applications ;-)
Adobe is scared of missing the boat and is now accepting AI images.
Scared for missing the boat? Well, if so, they could do as others, telling us that submit AI images is against their rules.
I think the opposite: Adobe is one step ahed other agencies, they know what is AI and the legal implications, so they decide to cross the river and give contributors ability to use the technology.
Personally I like this way, much better than tell us "you can't use it, only us can use the new technology"
The AI in Lightroom is just a tool to distinguish and select something in your photo (sky,object etc..) which is different in my opinion. But sure, they have somehow trained it to be able to do that. Does it also mean then they can automatically distinguish a JPG file that is real or made by AI? I wonder.
Your second point would mean that Adobe is not afraid of any lawsuit coming from whoever regarding copyrights or at least think that the claimer will have no leg to stand on.
I still think though that with this new AI tools, when it is getting near perfection, companies will not be using microstock agencies anymore or at least way less. Employees, in these big companies, are pretty competent enough to use those tools. They have to pimp your photo as well, don't they? So, then only small customers without that competency will remain but that is peanuts in the end. Editorial, obviously not included in that argument.
-
... They know much more than shutterstock or others. It would be really strange to think that they don't know and cannot disinguish an AI image from others....
how would they do that? the final result is a jpg
there is a comment in the iptc that says 'h OpenAI' but that is easily deleted.
-
The AI in Lightroom is just a tool to distinguish and select something in your photo (sky,object etc..) which is different in my opinion.
You miss that there are several "neural filters" in photoshop, wich cover different aspects of postproduction, and there is a "portrait generator" in coming soon tab that seems very interesting :)
They are using A LOT of AI potential.
Does it also mean then they can automatically distinguish a JPG file that is real or made by AI? I wonder.
I don't know for sure, of course it's only my opinion; anyway, for what I can see, they are building great experience in using AI in their software so I wouldn't be surprised to see they have something that can deeply analyze the image to understand if it's made by generative AI.
Something like AI that can recognize AI... not with metadata of course :)
Your second point would mean that Adobe is not afraid of any lawsuit coming from whoever regarding copyrights or at least think that the claimer will have no leg to stand on.
Once again it's only my opinion, and it's based on the experience that Adobe had already done, also from a legal point of view: as the AI in Lightroom and photoshop had been trained of course, so they know how and which terms they are legally using for these.
I still think though that with this new AI tools, when it is getting near perfection, companies will not be using microstock agencies anymore or at least way less. Employees, in these big companies, are pretty competent enough to use those tools.
Well, about this, I'm not so sure. It's hard to get good results with AI, and it takes time. For sure it takes more time than to find and buy a good ready-to-use image.
Let's think positive:
I can hazard a guess: AI will be the end for the 0,01 cent sales, but at the same time big sales for really good images designed and produced by humans will climb to new life with bigger money for who is able to produce them.
Maybe a dream, or a bet :) who knows, let's see
-
@Mat , please, let us know all the secrets of AI in Adobe ;D ;D ;D
-
@Mat , please, let us know all the secrets of AI in Adobe ;D ;D ;D
I will get right back to you on that!
-Mat
-
@mat
Wondering if you could shed some more light on the model/property release guidelines. I submitted some ai images which show a generic face. Sent a property release with these (following the guidelines) but they were rejected for the reason that they need a model release?
-
@mat
Wondering if you could shed some more light on the model/property release guidelines. I submitted some ai images which show a generic face. Sent a property release with these (following the guidelines) but they were rejected for the reason that they need a model release?
If a generative AI image is submitted with what could be perceived as a recognizable person in it, then you must submit a release no matter what.
If the person represented in the image is based on a real human, either via text prompt or image prompt, then a model release is required.
If the person is completely fictional and made up through generic text prompts not based on real people, then a property release is required.
@Dhav, will you please provide me with an image number as an example of a compliant file rejected in error? I would like to take a look. It's possible the release is non-compliant with our requirements.
This is a new and evolving process for us, so your patience is very much appreciated.
thank you,
Mat Hayward
-
Hi Mat,
I used a few generative ai images to turn them into png files, usually I knocked out the copy space area in the center, or on the side. They are usually simple watercolor or frame images/backgrounds.
Now they are sitting with an explanation that they need a property release.
For these files, I will probably just add a flat white background and resubmit as a normal illustration with generative ai label.
But I did add that they are generative ai illustrations and there are no people (real or fictional) in them.
Does this mean that going forward that every image gen ai I want to submit as png needs a property release?
Examples: 551777110 and 552447614
There are quite a few files that might be useful as pngs. But if I always need to add a release, then I will probably only do that for something special.
I state with every upload that these files are mine. Why is it different for pngs?
Thank you for always getting back to us so quickly.
-
I'd had AI images accepted before the new requirements. i submitted several images yesterday with the combined tags from the 2 posts and they were accepted.
still wondering if frequently photographed landmarks like taj mahal are acceptable, when no release needed for photographs
-
Just wanted to add, I really enjoy that I can process and upload my favorite ai images to Adobe. It is like an early christmas gift.
I don't think I will ever get my invested time and money for credits back, but it is just nice to see them accepted and out there.
And who knows, if you make a selection and create a special gallery, perhaps it will sell.
-
I recently had a conversation on Behance Live with the Adobe Stock content lead for vectors and illustrations, Shea Molloy about creating and submitting generative AI content for Adobe Stock. During the talk, we covered some of the basics of the content requirements that you have already seen. We did a deeper dive into some of the topics that people have asked for additional clarity on, such as release requirements, keyword requirements, etc.
If interested, you can watch the recording here: https://www.behance.net/videos/d5886b09-7fa7-4c0e-8a27-0d4b9451884a/Submitting-and-selling-generative-AI-images-on-Adobe-Stock-with-Mat-Hayward-and-Shea-Molloy (https://www.behance.net/videos/d5886b09-7fa7-4c0e-8a27-0d4b9451884a/Submitting-and-selling-generative-AI-images-on-Adobe-Stock-with-Mat-Hayward-and-Shea-Molloy)
Thanks,
Mat Hayward
-
So, as I understand this:
Shutterstock doesn't allow us to submit any AI generated images, but it will allow its customers to generate the AI images with our images (from Shutterstock database), and then compensate back to us if our images are being used to generate those new images.
Adobe Stock on the other hand, allows us to submit AI generated images. However, Adobe Stock doesn't want to take any responsibility of those images and want us to take full responsibility of it. As of now, for commoners like us, we can use Dall-E or Midjourney to create AI generated images, but we definitely doesn't own the rights of these images since it uses other people's images. BUT, some people do not care and take the risk to continue to upload images from this process because Adobe Stock has no ways to actually know if the uploader has the rights or not.
Is my understanding of the situations correct?
-
So, as I understand this:
Shutterstock doesn't allow us to submit any AI generated images, but it will allow its customers to generate the AI images with our images (from Shutterstock database), and then compensate back to us if our images are being used to generate those new images.
Adobe Stock on the other hand, allows us to submit AI generated images. However, Adobe Stock doesn't want to take any responsibility of those images and want us to take full responsibility of it. As of now, for commoners like us, we can use Dall-E or Midjourney to create AI generated images, but we definitely doesn't own the rights of these images since it uses other people's images. BUT, some people do not care and take the risk to continue to upload images from this process because Adobe Stock has no ways to actually know if the uploader has the rights or not.
Is my understanding of the situations correct?
Yes. If this is true. Why Adobe accepts AI content made with software that includes others works? I don’t see that in Adobe stock contribution terms or Adobe stock contribution terms has changed? This is serious question.
-
I really enjoyed the podcast, thank you. I also find it encouraging that you said that you are already seeing sales and lots of interest from buyers.
Like you say, this is absolutely not a get rich quick scheme. It takes me ages to come up with really good prompts and then hours to fix all the pixelmash glitches the ai creates. Even with what I would think are very, very simple prompts you get amazing f**** ups.
I still have the question why I need to add a property release, when I turn a gen ai file into a png, clearly label it as such and there are no people present in the image.
I have also uploaded a simple test video that I made for my own fun using one of my gen ai. Again I cleared labelled it as such and also added the file no of the base image in the description. Now we will see if it gets accepted ;)
Again, no faces or people in the image.
Loved how at the end producers are encouraged to add generative ai to all the content they already created this way. If you look at illustration content, or photos, from the last 18 months, all agencies have suspiciously similar content that looks ai produced.
Which is why the Adobe way is more honest. It is clearly labelled as such and the customer can decide for themselves if they might see any kind of legal risk in using this content.
Also important the emphasis to only use neutral prompts like bokeh,color,texture or futuristic, fantasy, dreamlike and not star wars, harry potter, banksy or tilmans or currently famous painters, musicians or cartoon characters. You can use „1930s detective“ as a prompt, but not Batman or Sherlock Holmes.
It is very obvious that Adobe put a lot of energy and resources into this project.
I hope it will be very successful. It is a very inspirational way to work, while the ai can be frustrating, it also comes up with some very creative ideas that can be quite inspiring.
But the credits add up…so I hope Adobe can put together a stunning collection for sales.
Have you considered taking some content as exclusive? I am sure a lot of producers would love to contribute and then this way we always know where the file came from.
-
Yes. If this is true. Why Adobe accepts AI content made with software that includes others works? I don’t see that in Adobe stock contribution terms or Adobe stock contribution terms has changed? This is serious question.
It has been explained several times. It doesn't work this way.
AI image DOES NOT include others works of any kind. It's a completely new and original image.
So this is not the point.
The legal point is if the images used TO TRAIN the AI engine was authorized and payed, or not. That's the legal point.
About author responsabilities this is always true for any agency and for any kind of content, so it's not new at all.
-
So, as I understand this:
Shutterstock doesn't allow us to submit any AI generated images, but it will allow its customers to generate the AI images with our images (from Shutterstock database), and then compensate back to us if our images are being used to generate those new images.
No.
Training the AI comes BEFORE the use, and they eventually ask and pay contributors one time only to train AI.
This is how it works.
No way to be payed back because AI images are original and new and does not include any part of the images used for training
-
@mat
Wondering if you could shed some more light on the model/property release guidelines. I submitted some ai images which show a generic face. Sent a property release with these (following the guidelines) but they were rejected for the reason that they need a model release?
If a generative AI image is submitted with what could be perceived as a recognizable person in it, then you must submit a release no matter what.
If the person represented in the image is based on a real human, either via text prompt or image prompt, then a model release is required.
If the person is completely fictional and made up through generic text prompts not based on real people, then a property release is required.
@Dhav, will you please provide me with an image number as an example of a compliant file rejected in error? I would like to take a look. It's possible the release is non-compliant with our requirements.
This is a new and evolving process for us, so your patience is very much appreciated.
thank you,
Mat Hayward
Hi @mat
Probably just me not understanding correctly but I submitted an ai portrait with a model release and that was rejected also. I think am example model release would be really helpful to see how we correctly fill them out.
-
@mat
Wondering if you could shed some more light on the model/property release guidelines. I submitted some ai images which show a generic face. Sent a property release with these (following the guidelines) but they were rejected for the reason that they need a model release?
If a generative AI image is submitted with what could be perceived as a recognizable person in it, then you must submit a release no matter what.
If the person represented in the image is based on a real human, either via text prompt or image prompt, then a model release is required.
If the person is completely fictional and made up through generic text prompts not based on real people, then a property release is required.
@Dhav, will you please provide me with an image number as an example of a compliant file rejected in error? I would like to take a look. It's possible the release is non-compliant with our requirements.
This is a new and evolving process for us, so your patience is very much appreciated.
thank you,
Mat Hayward
Hi @mat
Probably just me not understanding correctly but I submitted an ai portrait with a model release and that was rejected also. I think am example model release would be really helpful to see how we correctly fill them out.
I will need to see the release you submitted to identify the issue. Please email the file number to me via [email protected]
I understand there are still many questions about releases. The main thing to keep in mind is if your image contains what could be PERCEIVED to be a recognizable person (or property) then you must submit a release. If the person is totally fictitious, not based on an actual person in the written prompt, or an image based prompt, then a property release is what you need. If an actual person was used in any way, then a model release is needed.
There may be instances when you submit an image that doesn't contain a human, but you are asked for a release anyway. If that happens, simply attach a property release and resubmit.
You will find our release requirements at the following two links:
Model releases: https://helpx.adobe.com/stock/contributor/help/model-release.html
Property releases: https://helpx.adobe.com/stock/contributor/help/property-release.html
Thank you,
Mat Hayward
-
@MatHayward Just to clarify, does "generative ai" need to be one of the first 10 keywords?
I'm confused because I submitted 6 varied generative ai images on the same subject, with digital embellishments. Three were variations on a theme. Four were accepted (two of the variations) and two were put into the "Reminder" category telling me I could re-submit once I fixed the keywording and/or captioning. The captioning for one of the variations that was accepted and another put in "Reminder," were identical, but I moved generative ai from the 11th keyword to the 10th. All had "Illustration," "Generative AI," "Generative," and "AI" in the keywords and all noted they were generative ai in the captions, so I can't imagine anything else that could have been wrong with my captioning/keywording. I just fixed them today, so not sure if that cured the perceived problem.
Given that there have been requests for model releases when there isn't a human and other confusing requests on images submitted by long-time contributors who know to follow directions, I'm wondering how well reviewers have been trained on the new requirements? I understand when something's new like this there can be confusion on both sides.
_____________________________
I checked my images with Tineye & Google Images, and nothing even remotely similar showed up, even before I digitally edited them using my tablet.
I agree that AI is concerning - even frightening - for us creatives - just like microstock was for established professional stock photographers - but we can't change the march of technology whether it's the Model T or AI, there will be winners and losers with new technology. The best bet is to get on board and find a way to use it to our advantage.
I've been culling/reorganizing photos from the past 50+ years (personal stuff from childhood on, old family photos inherited from my parents, and my fine art and commercial work). I was struck by the changes in my lifetime (on so many levels! Those 80's shoulder pads!), as well as by how badly my color photos from as late as the 1980s have discolored even when kept in archivally sound environments. I put my faves aside to scan, knowing that, with the white balance eye dropper in Photoshop, they'll look better than new, ready to upload to my favorite photo lab, for delivery to my door overnight.
We can't pick and choose the tech we want.
-
@MatHayward Just to clarify, does "generative ai" need to be one of the first 10 keywords?
I'm confused because I submitted 6 varied generative ai images on the same subject, with digital embellishments. Three were variations on a theme. Four were accepted (two of the variations) and two were put into the "Reminder" category telling me I could re-submit once I fixed the keywording and/or captioning. The captioning for one of the variations that was accepted and another put in "Reminder," were identical, but I moved generative ai from the 11th keyword to the 10th. All had "Illustration," "Generative AI," "Generative," and "AI" in the keywords and all noted they were generative ai in the captions, so I can't imagine anything else that could have been wrong with my captioning/keywording. I just fixed them today, so not sure if that cured the perceived problem.
No, you do not need to list generative AI in the top ten keywords. It can be number 49 if that is your preference. I don't see any way that would have influenced the decision to put the file into the reminder category. Usually that happens if there are irrelevant or inappropriate keywords. Please send me the file number and I will take a look. There is a learning curve for all of us, so the more data I can review in these early days, the better. Email [email protected]
thanks,
Mat Hayward
-
Hi Mat,
I understand why in the podcast you said we should never use 3d render, oil painting, watercolor in our description and keywords, because it was not created this way. It is probably a hardcore legal necessity for the project.
It was then suggested to look for wordings that might inflect 3d render without being 3d render, like low poly.
But if I put on my customer hat and I am looking for „Beautiful flowers in a vase oil painting“, I personally wouldn‘t care if it was gen ai or a real photo of an oil painting.
And if I look specifically for generative ai, then just adding the combo „beautiful flowers in a vase generative ai“, will not give me the results I am looking for. Because it will also return everything that was used with the flower prompts, i.e. images done in pixel art, watercolor, woodcuts,impressionism etc…
Adobe could create a new list of keywords for producers and buyers that we can use instead of the actual -
„oil color - generative ai colors looking like liquid fossil fuel paint“etc…
but this will become complicated very quickly.
My suggestion - why not let us use the following:
generative ai, ai,generative as keywords and then
„oil painting style“, „3d render style“, „woodcut style“ and always only to be used in combination with generative ai in titles, descriptions and keywords.
Customers could then be informed how to search visually for what they need with a very simply tweak of the searches they are used to.
If you have millions of gen ai files, how else will the customer find what they need? How many word variations of oil painting can they think of to find content? And will their ideas match what the producers are using to describe their content without describing their content…?
Would that work from a legal perspective?
And if 3d renders specifically are so important, why not make it a seperate choice upon upload - photo, illutsration,3d render. Wouldn‘t that make it even easier to seperate inside the collection and give customers an easier way to search?
Just an idea, hope it is not a silly suggestion.
But you will have 30 million files at some point, because working with gen ai is extremely popular as a hobby already.
-
@MatHayward Just to clarify, does "generative ai" need to be one of the first 10 keywords?
I'm confused because I submitted 6 varied generative ai images on the same subject, with digital embellishments. Three were variations on a theme. Four were accepted (two of the variations) and two were put into the "Reminder" category telling me I could re-submit once I fixed the keywording and/or captioning. The captioning for one of the variations that was accepted and another put in "Reminder," were identical, but I moved generative ai from the 11th keyword to the 10th. All had "Illustration," "Generative AI," "Generative," and "AI" in the keywords and all noted they were generative ai in the captions, so I can't imagine anything else that could have been wrong with my captioning/keywording. I just fixed them today, so not sure if that cured the perceived problem.
No, you do not need to list generative AI in the top ten keywords. It can be number 49 if that is your preference. I don't see any way that would have influenced the decision to put the file into the reminder category. Usually that happens if there are irrelevant or inappropriate keywords. Please send me the file number and I will take a look. There is a learning curve for all of us, so the more data I can review in these early days, the better. Email [email protected]
thanks,
Mat Hayward
Thanks Mat - I didn't send them along since they were accepted today.
-
I just had everything declined, over 20 files from different subject matters and including several pngs.
Since most of my uploads were accepted before, and just two days ago I had 31 accepted in one batch, this is quite frustrating.
Especially since this week I already got my first gen ai sales, one file sold today was uploaded just 48hours ago.
Since I haven't changed the way I work, could Adobe please give us some guidance of what they want.
I know people say, gen ai is not work, but it is actually more expensive because of the dalle fees, at least in my case. And of course I am not perfect with my prompts.
So I am taking a break and will focus on other things.
-
I just had everything declined, over 20 files from different subject matters and including several pngs.
Since most of my uploads were accepted before, and just two days ago I had 31 accepted in one batch, this is quite frustrating.
Especially since this week I already got my first gen ai sales, one file sold today was uploaded just 48hours ago.
Since I haven't changed the way I work, could Adobe please give us some guidance of what they want.
I know people say, gen ai is not work, but it is actually more expensive because of the dalle fees, at least in my case. And of course I am not perfect with my prompts.
So I am taking a break and will focus on other things.
What was the specific rejection reason?
-
quality problems
but that seems to be a general catch all.
I process everything the same way and just 48h ago everything was accepted. I have been uploading nearly every day for a month and most files go in. So this is very abrupt.
Two weeks ago I had 10 files declined in one go, but they were all cartoon style drawings and I interpreted that this is not the type of content they want, since I also cannot see much of it in the gen ai collection. Might be a more difficult legal issue or simply not the content they are looking for.
But this was a batch similar in themes that were all accepted. So far I have uploaded 202 files, including the ones rejected now and I have 165 accepted.
I have processed over 3000 files on dalle, mainly because I am trying to learn about working with dalle, it is quite different from the nightcafe creator I was using before.
So I am only sending a tiny portion of what I experiment on to Adobe. Most prompts or concepts simply don't work as intented. I am getting better, but it will probably take me at least a year to get to the level I would like to reach. Maybe even 2.
I am sure the pro illustrators have much better success rates, but I am learning something new. It is expensive, but a. lot of fun. Actually sometimes the mistakes are the most interesting.
For me this is also a reminder never to fall too much in love with any agency. We are running a business and it is important to spread your attention to the whole market.
I was just enjoying it too much and this week I got the first 5 sales :(
I guess it is back to normal photography and objects on white. Boring, but it pays the bills. Will still do gen ai, just upload it elsewhere in the meantime.
-
quality problems
but that seems to be a general catch all.
I process everything the same way and just 48h ago everything was accepted. I have been uploading nearly every day for a month and most files go in. So this is very abrupt.
Two weeks ago I had 15 files declined in one go, but they were all cartoon style drawings and I interpreted that this is not the type of content they want, since I also cannot see much of it in the gen ai collection. Might be a more difficult legal issue or simply not the content they are looking for.
But this was a batch similar in themes that were all accepted. So far I have uploaded 202 files, including the ones rejected now and I have 165 accepted.
I have processed over 3000 files on dalle, mainly because I am trying to learn about working with dalle, it is quite different from the nightcafe creator I was using before.
So I am only sending a tiny portion of what I experiment on to Adobe. Most prompts or concepts simply don't work as intented. I am getting better, but it will probably take me at least a year to get to the level I would like to reach. Maybe even 2.
I am sure the pro illustrators have much better success rates, but I am learning something new. It is expensive, but a. lot of fun. Actually sometimes the mistakes are the most interesting.
For me this is also a reminder never to fall too much in love with any agency. We are running a business and it is important to spread your attention to the whole market.
I was just enjoying it too much and this week I got the first 5 sales :(
I guess it is back to normal photography and objects on white. Boring, but it pays the bills. Will still do gen ai, just upload it elsewhere in the meantime.
I appreciate the feedback on your experience. I took a look at your rejected content and I'm sorry to report I think the moderators got it right. There were some funky crops and general quality issues that maybe came from upsizing? I recommend trying out some different software programs to see if you come up with better results.
Take a look at the collection as it sits at Adobe Stock now to get an idea of the level of competition. Before uploading, look at each file through the lens of an impartial content curator. If you can do that, you'll save a lot of time and effort. by submitting only high quality content with a high probability of strong sales. I agree that it takes a lot of work to create sellable stock content with this new technology. There is a steep learning curve that we are all trying to figure out as our portfolios continue to grow.
I wish you the best of luck with your future submissions!
-Mat Hayward
-
Hi Mat,
thank you for taking the time to look at my files.
I really want to improve, so I will try to learn more. And if it really is technical quality and not the images themselves, that is a solvable problem.
In the meantime, I will do more boring stuff.
Have a great holiday.
eta
just had this recommended, will try this and others and then hopefully I will get a workflow with better quality.
sorry for whining. :)
https://www.topazlabs.com/gigapixel-ai?fbclid=IwAR2k9cRvEGQFVM0UmXSeHIeAMlPmg0xEbfTIC0BPj8hRcy3K5UAPeY9FQpQ (https://www.topazlabs.com/gigapixel-ai?fbclid=IwAR2k9cRvEGQFVM0UmXSeHIeAMlPmg0xEbfTIC0BPj8hRcy3K5UAPeY9FQpQ)
eta2
just wanted to share, last December I made a measly 11 dollars. I basically hadn't uploaded in years, so I deserved that. Now I have added around 300 files in 3 months, and this month is already 10 times more than last December, 120 at the moment. Currently have 1750 files. This weeks position around 5600, up from 20k a few weeks ago.
Uploading to Adobe works!
Next target: 5000 useful files. Will probably take 2 years.
-
This stuff here is crazy. Truly movie level quality.
https://www.facebook.com/groups/officialmidjourney/permalink/476721641286072/ (https://www.facebook.com/groups/officialmidjourney/permalink/476721641286072/)
Next level will be animating this and bringing them into your ai generated videos.
-
@ MatHayward
Hi Mat,
sorry but I have an issue/problem with property release and I would like to ask for clarification.
From Adobe FAQ
"You may submit the same release for content created for the same project or within the same series, which includes works with the same model or similar themes created on or about the same date"
(https://helpx.adobe.com/stock/contributor/help/generative-ai-content.html)
And following this I send series of people with appropriate releases for each series /theme
One time all images was accepted
Two times all images was pushed in review for "property release issue", even the same images used for references in the releases.
Can you please clarify how to use the property release for generated AI people?
If you don't mind I'm sending you in pvt the image numbers to take a closer look.
It seems hard tpo understand exactly what to do, if a property release for each image, for variations, for themes, and why sometimes that are accepted sometimes refused.
Many Thanks
-
@ MatHayward
Hi Mat,
sorry but I have an issue/problem with property release and I would like to ask for clarification.
From Adobe FAQ
"You may submit the same release for content created for the same project or within the same series, which includes works with the same model or similar themes created on or about the same date"
(https://helpx.adobe.com/stock/contributor/help/generative-ai-content.html)
And following this I send series of people with appropriate releases for each series /theme
One time all images was accepted
Two times all images was pushed in review for "property release issue", even the same images used for references in the releases.
Can you please clarify how to use the property release for generated AI people?
If you don't mind I'm sending you in pvt the image numbers to take a closer look.
It seems hard tpo understand exactly what to do, if a property release for each image, for variations, for themes, and why sometimes that are accepted sometimes refused.
Many Thanks
Hi Derby,
Thanks for sharing the file numbers with me, it's very helpful. As mentioned separately, I've forwarded those to our moderation team for a re-review as it's possible a mistake was made. If not, I'll be sure to let you know the specific reason.
I'll be mostly offline through the end of next week but will give you an update before then if I hear back.
Thank you!
Mat Hayward
-
Thanks Mat for your help!
-
fwiw my gen ai are selling. Considering that most content has been online for less than 3 weeks, this is quite encouraging.
Happy New Year everyone!
-
Happy New Year to all!
I might submit some of my Midjourney results this year, so I was searching the Adobe database to learn more.
However, I found images that were clearly Midjourney generated, without the "AI generated" label - I have no problem with that, but I do have a question : in search mode, how to filter only model released images ?
In the left column (filtering), I can select "illustrations" and "with people", but is there anyway to find out if an image is model released? This would enable a buyer to know if the "AI generated person" in the image is depicting a real person or a fictitious one?
Example : if I search for "child reading fairy tale bear" + illustration filter :
- "with people" enabled gives me 80% illustrations WITHOUT people (lots of teddy bears, books etc.)
- "without people" enabled gives me 15% images with AI generated people (some without the AI tags)
So how does a buyer know if the person in the illustration depicts a real person or not? and how can a buyer find out if a model release has been submitted or not? (of course, this last question has nothing to do with AI generated content).
If "without people" means "no model release", then clearly quite a few early submissions got through without a release. But if "with people" means there is a MODEL release, then why do we get so many bear/book illustrations ?
-
Wonderful, Adobe & Mat! This new policy is totally in line with abolishing the bonus code for 2023. Screw human artists! Bring on the robots!
I don't think I'll bother with photography any more. I'll get an army of robots and train them to blab text commands to robotic AI software night and day. These smart robots will vacuum pics generated by other robots, re-order pixels and create brilliant new pictures. (some say that's crap but they're only jealous, and live under a rock or something) Soon I'll have a portfolio of 100 Million brilliant AI images on Adobe Stock. I will be so rich! I will buy a villa in some place tropical, heck, I will buy an island of my own! I will buy a luxury plane that makes royalty, presidents and celebs jealous!
Thank you Adobe, you made my day. Screw artists, Robots are our future.
-
And does anyone even know what programs can generate these images? And what are the PC requirements?
I agree - I'd be interested in checking out the software too. Which application(s) are people using to do this? Thanks! Please feel free to PM or post here, thanks!
-
Wonderful, Adobe & Mat! This new policy is totally in line with abolishing the bonus code for 2023. Screw human artists! Bring on the robots!
I don't think I'll bother with photography any more. I'll get an army of robots and train them to blab text commands to robotic AI software night and day. These smart robots will vacuum pics generated by other robots, re-order pixels and create brilliant new pictures. (some say that's crap but they're only jealous, and live under a rock or something) Soon I'll have a portfolio of 100 Million brilliant AI images on Adobe Stock. I will be so rich! I will buy a villa in some place tropical, heck, I will buy an island of my own! I will buy a luxury plane that makes royalty, presidents and celebs jealous!
Thank you Adobe, you made my day. Screw artists, Robots are our future.
I'm old enough to remeber quite similar comments in the 90's when digital cameras comes out for large consumer audience.
Using these words means you've really never tested the AI tools. It's super exciting; but it's absolutely not easy to generate real good images.
It's not easy to use.
It costs money and it costs a lot of time.
It's relatively easy to produce in random way some incredible images, but it's abosultely hard to manage the tool to produce series of images with steady style.
Of course it will be easier in the next future; and it will become also cheaper.
But, as for photography, you always have to start from a good idea, and you also need cultural, and technical, skills.
Back to Adobe: I think at the contrary that giving contributors the option to produce (and sell!) AI images is a great opportunities to partecipate in production; at the opposite, other agencies like SS seems that will cut contributors, pushing their clients to produce images by ourselves. In my opinion this last way is absolutely wrong, starting from the idea that anyone have clear idea and skill to create any image they need.
One last example I know very well: in movie production I come from times when editing was made with film, scissor, glue... with big pain and time lost. Digital non linear editing completely changed the game but directors always need good editor to work, also to make choices. To have a digital machine that gives you thousand option in a single second doesn't mean that you don't need human time anymore to make your decisions and choices. Machine is fast, human mind has its own timing.
Finally, AI development is impossible to stop: refusing it and complaining about AI is totally useless, it will be the next future for all of us
-
And does anyone even know what programs can generate these images? And what are the PC requirements?
I agree - I'd be interested in checking out the software too. Which application(s) are people using to do this? Thanks! Please feel free to PM or post here, thanks!
i've been using DALL-E & pleased w the results when i use it for illustrations - it's expanded my reach as it's an area i could approach. you get 15 free tries each month
-
...
Using these words means you've really never tested the AI tools. It's super exciting; but it's absolutely not easy to generate real good images.
It's not easy to use....
Finally, AI development is impossible to stop: refusing it and complaining about AI is totally useless, it will be the next future for all of us
First part. LOL. 2 points re. that
1. You know anyone can log into Midjouney and see exactly how "hard" it is in real time don’t you? (you get to see the prompts and the images produced in real time, highly recommend if you want to know how seriously to take these comments). Alternatively watch a YouTube video or two.
2. You know we can still see older work in the portfolios of people claiming this is just a new tool in an “artists” toolbox? (and how they instantly made a quantum leap not only in craft but also in composition, artistic vision etc.)
These arguments are silly. The evidence is right there in the open, at least until Midjouney comes off a discord server and artists delete their older work so we don’t get to witness the process anymore.
Second part, absolutely agree. It wont kill off art but commercial work illustrators will dry up very quickly.
-
First part. LOL. 2 points re. that
1. You know anyone can log into Midjouney and see exactly how "hard" it is in real time don’t you? (you get to see the prompts and the images produced in real time, highly recommend if you want to know how seriously to take these comments). Alternatively watch a YouTube video or two.
2. You know we can still see older work in the portfolios of people claiming this is just a new tool in an “artists” toolbox? (and how they instantly made a quantum leap not only in craft but also in composition, artistic vision etc.)
These arguments are silly. The evidence is right there in the open, at least until Midjouney comes off a discord server and artists delete their older work so we don’t get to witness the process anymore.
I know I can copy prompts; but learning how to write a prompt to obtain your unique image it's completely different job.
Yes of course you can copy, and it's far from necessary to copy. You can also obtain hundred of nice images for (let's say) valentine's day with really simple inputs: copy prompts it's not an issue if you can have a good image only asking "valentine's day".
But: creating an original different work you are thinking about is completely different: it takes time and several inputs; and midjourney it's not free, and not cheap at all (you pay for a year more or less the same than Adobe creative cloud subscription).
When you finally have a good image (or series of), that's not enough: upscaling is a pain, takes more time, and once again it's not free, you have to pay software.
Finally you can do title and keyword.
Are you sure that so many people will pay so much to obtain few money flooding agency with hundred of similar images? I don't think it will go this way.
In my opinion, at the beginning (now) there will be the flood; but soon many people will realize that earn money from AI it's not automatic, not fast, not easy.
EDIT: I would like to add an example from my real last days.
I went in a music studio to shoot musical instruments, interiors, details, various eqwuipment and so on... It takes three hours to produce about 100 good photos.
Then three hours more for post production, and I have 100 photos ready to go; total six hours of work.
100 good images with AI with same subject... it will take not less than three/four days of full immersion job in front of computer. Of course I'm talking about GOOD images, not one hundred of random musical instruments.
(keywording is the same for both works so it doesn't matter)
This is what I mean when I say "it's hard to work on it"
-
I wasn't referring to copying prompts. I was referring to seeing how prompts are used and how much work there is to learn and refine those prompts to get the desired results.
The other barriers to entry you reference, like the upscaling, are very temporary, and clearly not that much of a barrier even now when you see how many high quality images a lot of one man bands are uploading.
-
Wonderful, Adobe & Mat! This new policy is totally in line with abolishing the bonus code for 2023. Screw human artists! Bring on the robots!
I don't think I'll bother with photography any more. I'll get an army of robots and train them to blab text commands to robotic AI software night and day. These smart robots will vacuum pics generated by other robots, re-order pixels and create brilliant new pictures. (some say that's crap but they're only jealous, and live under a rock or something) Soon I'll have a portfolio of 100 Million brilliant AI images on Adobe Stock. I will be so rich! I will buy a villa in some place tropical, heck, I will buy an island of my own! I will buy a luxury plane that makes royalty, presidents and celebs jealous!
Thank you Adobe, you made my day. Screw artists, Robots are our future.
I'm old enough to remeber quite similar comments in the 90's when digital cameras comes out for large consumer audience.
Using these words means you've really never tested the AI tools. It's super exciting; but it's absolutely not easy to generate real good images.
It's not easy to use.
It costs money and it costs a lot of time.
It's relatively easy to produce in random way some incredible images, but it's abosultely hard to manage the tool to produce series of images with steady style.
Of course it will be easier in the next future; and it will become also cheaper.
But, as for photography, you always have to start from a good idea, and you also need cultural, and technical, skills.
Back to Adobe: I think at the contrary that giving contributors the option to produce (and sell!) AI images is a great opportunities to partecipate in production; at the opposite, other agencies like SS seems that will cut contributors, pushing their clients to produce images by ourselves. In my opinion this last way is absolutely wrong, starting from the idea that anyone have clear idea and skill to create any image they need.
One last example I know very well: in movie production I come from times when editing was made with film, scissor, glue... with big pain and time lost. Digital non linear editing completely changed the game but directors always need good editor to work, also to make choices. To have a digital machine that gives you thousand option in a single second doesn't mean that you don't need human time anymore to make your decisions and choices. Machine is fast, human mind has its own timing.
Finally, AI development is impossible to stop: refusing it and complaining about AI is totally useless, it will be the next future for all of us
I totally agree, make decisions and choices is the difference ;)
-
I am doing lots of test searches and looking at new content on Adobe for gen ai. It seems that fresh content is mostly 3d render style and a lot of photorealism, and mostly from Midjourney. I don‘t see a lot of charcoal drawings, sketches, oil paintings or abstracts. There is less variety in style.
Maybe this is the style the customers mostly buy? There is no question it will be the dominant ai look for the next few years, the way instagram filters dominated photos destined for social media and blogs. But you also need variety.
I am sure there will soon be a midjouney app, or an Adobeapp working with Midjourney.
Upsampling with Gigapixel ai is certainly much better than upsampling with photoshop. Thank you for all the tips.
-
I don‘t see a lot of charcoal drawings, sketches, oil paintings or abstracts. There is less variety in style.
Other than actually use it as an art print to hang on a wall or maybe greeting cards there isn't really much usage to customers for charcoal drawings, oil paintings and the likes. I am not saying there was no market for this, but it's a limited one.
The typical midjourney style AI images though that look like computer graphic art? The possibilities are almost endless. You can also use them to hang on a wall or as a postcard like with charcoal or oil paintings, but you can also use them for book covers (fantasy and Sci-Fi for example have drawn covers 90% of the time) , RPG books and advertisement, computer and online game art and advertisement, shirts, CD covers (especially in the metal genre art is more poular on covers than photos) and so on. It just makes sense to use a style that has a much wider range of application, thus more sale potential.
-
I think the Midjourneylook is also very seductive and beautiful. So everyone is just joining in the fun. It will become even stronger once the first Midjourney app is released and the general public can start playing as well.
But all ais can do a lot more than just one kind of style.
In my other life in business with an engineering company, the warm hearted fantasylore vibe with typical midjourney colors would not be useful. But Midjourney can do a lot more, just people are currently not uploading that.
So even if you do render styles and photorealism, there is a lot more variety possible than I am currently seeing in the uploads.
If you sort any search by photos, you will get a huge variety in styles and ideas. But If I do a search for anything in generative ai and sort by new it looks like 90% is content created with the very typical Midjourney look. At the moment the new gen ai looks more like a "collection" and not like a content type.
Even without using other art techniques like watercolor or cartoons, just within renders and photorealism, there is a lot more you can do.
I think especially for business/engineering there is a wide open field for a more neutral look.
Other ais can also do some very interesting stuff.
Just my 2 cents with my buyer hat. Obviously the Midjourneylook will have a crazy success for 2 years. But what happens then?
Here are some prompt results for stable diffusion.
https://mpost.io/best-100-stable-diffusion-prompts-the-most-beautiful-ai-text-to-image-prompts/ (https://mpost.io/best-100-stable-diffusion-prompts-the-most-beautiful-ai-text-to-image-prompts/)
And this article
https://www.arxiv-vanity.com/papers/2209.11711/ (https://www.arxiv-vanity.com/papers/2209.11711/)
I am collecting articles around ais.
-
Hi everyone,
A quick update to let you know we have added a new generative AI checkbox in the Adobe Stock Contributor Portal. In the upper right corner, you will see a checkbox that says: "created with generative AI tools." Select this generative AI checkbox if any part of your image was created using generative AI software. Please note that other than this additional step, the submission guidelines have not been changed:
-Submit each file as an illustration, even if it resembles a photograph
-Include "generative AI" in your title and as an individual keyword.
-Add "generative" and "AI" as additional individual keywords.
I recommend if you are submitting multiple generative AI images, that you select all in the portal and take the appropriate steps listed above before completing your indexing process and submitting. Generative AI is evolving quickly and we are grateful to everyone who has taken the guidelines to heart and submitted fully compliant content. Please refer to the submission guidelines here for any questions or feel free to ask me here.
Thank you,
Mat Hayward
-
-Submit each file as an illustration, even if it resembles a photograph
-Include "generative AI" in your title and as an individual keyword.
-Add "generative" and "AI" as additional individual keywords.
Thank you,
Mat Hayward
Hi Mat,
Thanks, that is clear to me, but I still have a question about model releases.
Is there a difference in the way I submit the following images :
- a photo (non AI) of a real person photoshopped into an AI background + model release signed by the model
- an image fully created in AI, with a non-existing person created in AI + model release signed by me
Buyers never see the model releases - they cannot even see if a model release exists - so my question is : how does a buyer know if the person in the image is real or not? Is there something the artist can/should do to let de buyer know ?
-
Thank you for the checkbox, very helpful.
Do we still have to avoid using the names of the art technique used, i.e. we should still NOT include watercolor style, Oil Painting style, 3d render style?
I am doing test searches and without art techniques all I can do in searches is color and subject or maybe copy space.
-
-Submit each file as an illustration, even if it resembles a photograph
-Include "generative AI" in your title and as an individual keyword.
-Add "generative" and "AI" as additional individual keywords.
Thank you,
Mat Hayward
Thanks for the question, Anyka. If any part of the image was created with generative AI software, you must include the metadata, check the box, and submit as an illustration. Please be sure to reference the submission guidelines and FAQ for additional information.
-Mat Hayward
Hi Mat,
Thanks, that is clear to me, but I still have a question about model releases.
Is there a difference in the way I submit the following images :
- a photo (non AI) of a real person photoshopped into an AI background + model release signed by the model
- an image fully created in AI, with a non-existing person created in AI + model release signed by me
Buyers never see the model releases - they cannot even see if a model release exists - so my question is : how does a buyer know if the person in the image is real or not? Is there something the artist can/should do to let de buyer know ?
-
Thank you for the checkbox, very helpful.
Do we still have to avoid using the names of the art technique used, i.e. we should still NOT include watercolor style, Oil Painting style, 3d render style?
I am doing test searches and without art techniques all I can do in searches is color and subject or maybe copy space.
What is most important is that you do not imply the image is something other than generative AI.
Thanks Cobalt,
Mat Hayward
-
That is very helpful, thank you.
I don‘t think there is a big market for these, but I am having fun.
https://stock.adobe.com/de/stock-photo/id/562127482
I used to do sketches and charcoal as a hobby, now I can do it faster (and better) with ai.
But they need a proper description.
-
I have generated a few files with WOMBO Dream AI Art Tool app on the iphone.
It's not clear if I can submit these images to adobe stock.
Can you please tell me if I can.
I don't want to infringe any copyright and I want to be sure.
This is what I found about copyright policy on the site of the software I would like to use:
Creators have full freedom over their artwork and are free to distribute or share them for personal, entertainment or educational reasons. For commercial use, please credit or provide attribution to WOMBO for the generation in some way. We recommend using the trading card output if possible, but a mention in a credits page is fine too.
So I can use the generated files for commercial purpose but still I have to provide attribution. But in this case I don't know how to provide attribution to them and so I don't know if I can go on and submit those files to you.
Can you please help me?
-
Why don‘t you write to Wombo support and ask for a written statement that you can sell your images on Adobestock?
Nobody here can give you legal advice, only the ai company can do that.
-
or use an AI generator that gives you complete ownership w no
need to give credit to the app (after all, we dont need to credit PS when we use their AI filters or other smart tools)
-
I asked to WOMBO and this is their answer:
Hi Giovanni - Creators have full freedom over their artwork and are free to distribute or share them for personal, entertainment or educational reasons.
For commercial use, please credit WOMBO Dream for the generation in some form. We recommend using the trading card output if possible, but a mention in the credits is also fine!
So I haven't solved my problem.
Anyway, what companies allow you to use their AI generated images?
-
I just received another email from them. after telling them I have not the possibility of crediting them:
In this case, feel free to sell your creations without crediting us!
Is this answer sufficient? Can I start uploading the generated files?
-
Is this answer sufficient? Can I start uploading the generated files?
Reading the Adobe terms
is your responsability, submitting generative AI content, to own all the necessary rights.
So, it's up to you to consider (or not) enough the Wombo mail, no one else could answer you
From the FAQ: https://helpx.adobe.com/stock/contributor/help/generative-ai-content.html
Ensure you have the appropriate rights to submit.
You must have all the necessary rights to submit generative AI illustrations to Adobe Stock for licensing and use as described in our contributor terms (e.g., broad commercial use). You must review the terms of any generative AI tools that you use to confirm that this is the case before you submit any AI-generated content.
Do: Read the terms and conditions for generative AI tools that you use to ensure that you have the right to license all generative AI content that you submit to Adobe Stock under the contributor terms. For example, you cannot submit any content if you are not permitted to license it for commercial purposes.
-
The problem is it's not completely clear and I don't want to my adobe account. Are content generated from Dall-E and Midjourney ok for commercial use and to upload to adobe stock?
-
While there are ethical questions about the generators themselves (because they did not get consent from the creators to create their 'trained' models) - from the tos of midjourney - if you pay for their service, you have general commercial terms. from that other site you mentioned - they basically said yes, you have permission as well. dall-e - haven't seen their tos so don't know.
-
I found this on the dalle page, so I use dalle and upload to Adobe without problems.
https://help.openai.com/en/articles/6425277-can-i-sell-images-i-create-with-dall-e
Can I sell images I create with DALL·E?
Commercial use of DALL·E
Written by Raf
Updated over a week ago
Subject to the Content Policy and Terms, you own the images you create with DALL·E, including the right to reprint, sell, and merchandise – regardless of whether an image was generated through a free or paid credit.
But the majority of great content on Adobe comes from Midjourney, so there must be a legal announcement somewhere.
I am not a lawyer, use any info you find at your own risk.
I do pay dalle for credits. I don‘t want to upload anything for commercial use without paying for the usage.
-
I just hope that the customers could be opt out AI products in the search result 8)
-
I just hope that the customers could be opt out AI products in the search result 8)
why would they if AI images meet their needs?
-
What about hybrid images? Say a photo of a hand holding a phone, displaying an ai generated image. Or a generated room interior, with a real framed photo hanging on the wall. There could be infinite ways to combine camera photos with 3D renders, ai images, and even hand drawn illustrations.
-
I just hope that the customers could be opt out AI products in the search result 8)
why would they if AI images meet their needs?
Because, at my thought, AI give the same taste on every things. And this is in-aesthetic. I think we all would need, in our life, something more authentic. 8)
-
I just hope that the customers could be opt out AI products in the search result 8)
why would they if AI images meet their needs?
And if it doesn't meet their needs? It is already hard to find something specific on stock sites, now on top of that they have to search through thousands of AI images.
-
I just hope that the customers could be opt out AI products in the search result 8)
why would they if AI images meet their needs?
Because so many people are copying the same prompts, you get endless very, very similar looking files. A lot more similar than in normal type of copying. This will get much worse, when Midjourney releases their app to the world.
So my buyers hat would like to be able to opt in or out of gen ai content.
I might also decide I want to only see gen ai files, for a specific project.
Give customers the choice and everyone is happy.
-
I just hope that the customers could be opt out AI products in the search result 8)
why would they if AI images meet their needs?
And if it doesn't meet their needs? It is already hard to find something specific on stock sites, now on top of that they have to search through thousands of AI images.
as opposed to the millions they already have to wade thru?
why do you assume that AI won'ty supply 'someth ing specific'? and if they don't they are no threat to existing portfolios -- you can't claim AI is replacing traditional and that AI isnt filling a need! (actually, i guess you can claim that, but it doesnt make much sense)
-
I just hope that the customers could be opt out AI products in the search result 8)
why would they if AI images meet their needs?
And if it doesn't meet their needs? It is already hard to find something specific on stock sites, now on top of that they have to search through thousands of AI images.
as opposed to the millions they already have to wade thru?
why do you assume that AI won'ty supply 'someth ing specific'? and if they don't they are no threat to existing portfolios -- you can't claim AI is replacing traditional and that AI isnt filling a need! (actually, i guess you can claim that, but it doesnt make much sense)
If buyer thinks he will find something he needs in AI images, he could turn it on, if he doesn't, he could turn it off. Simple as that. I am not claiming AI is replacing traditional, I don't remember saying something like that.
-
AI eventually will replace the photographer, the contributor and the agency. And the customer will get his/her images for free. It is matter of time.
-
AI eventually will replace the photographer, the contributor and the agency. And the customer will get his/her images for free. It is matter of time.
You can already download millions (billions?) of files for free legally from many different platforms.
Why do agencies still exist?
-
AI eventually will replace the photographer, the contributor and the agency. And the customer will get his/her images for free. It is matter of time.
You can already download millions (billions?) of files for free legally from many different platforms.
Why do agencies still exist?
True, but also sales are not even close to what they used to be. It is increasingly getting harder because all these reasons, free images, phone images, lower royalties, higher greed of agencies, now ai and who knows what else.
-
Yes, there is a flood of images and many files actually have a very long shelf life. I am amazed that content from 2005 is still selling inspite of all the new uploads.
Ai will add another stress factor, but it is also an opportunity. IMO this is better than competing with a flood of free files or endless duplicates by the newbies sorting our ports by downloads somewhere and copying what works.
One of the reasons I avoided Dreamstime for a long time.
But Ai gives completely new opportunities for content, especially for concepts that might be difficult to shoot or simply save you time by visualizing your next shoot.
I think it is a great tool. For now.
-
AI eventually will replace the photographer, the contributor and the agency. And the customer will get his/her images for free. It is matter of time.
You can already download millions (billions?) of files for free legally from many different platforms.
Why do agencies still exist?
Companies with legal departments (like Disney - ABC) require employees and contributors (often graphic designers or journalists) to have releases / permission for everything including downloaded visual content. The unusual high royalty at Shutterstock is probably Shutterstock charging for the inconvenience of having their licensing terms expanded and re-written.
-
David Holz, midjourney CEO, said less than an hour ago that he doesn't like midjourney-generated images being put up for sale through stock agencies. He said that he is seriously considering banning the sale of midjourney-generated images through stock agencies.
-
David Holz, midjourney CEO, said less than an hour ago that he doesn't like midjourney-generated images being put up for sale through stock agencies. He said that he is seriously considering banning the sale of midjourney-generated images through stock agencies.
Where?
-
David Holz, midjourney CEO, said less than an hour ago that he doesn't like midjourney-generated images being put up for sale through stock agencies. He said that he is seriously considering banning the sale of midjourney-generated images through stock agencies.
I can't find any source for this.
And it would be a complete nonsense, if real: why in the earth Midjourney would donate this market to dozens of competitor?
Also from legal point of view it seems very hard to say "You have commercial rights BUT not for this specific market". David Holz cannot "ban" that market, at the contrary the market can "ban" midjourney, that has been already done by SS and Getty.
By the way... it's a strange world and anything it's possible :) I would like to read the original source with the words of David Holz
-
David Holz, midjourney CEO, said less than an hour ago that he doesn't like midjourney-generated images being put up for sale through stock agencies. He said that he is seriously considering banning the sale of midjourney-generated images through stock agencies.
I can't find any source for this.
And it would be a complete nonsense, if real: why in the earth Midjourney would donate this market to dozens of competitor?
Also from legal point of view it seems very hard to say "You have commercial rights BUT not for this specific market". David Holz cannot "ban" that market, at the contrary the market can "ban" midjourney, that has been already done by SS and Getty.
By the way... it's a strange world and anything it's possible :) I would like to read the original source with the words of David Holz
It depends on the company's Terms. Take for example the two biggest marketplaces for 3D models - Turbosquid and CGTrader.
Turbosquid doesn't allow images, generated from their models to be sold on stock markets, but CGTrader does.
-
It depends on the company's Terms. Take for example the two biggest marketplaces for 3D models - Turbosquid and CGTrader.
Turbosquid doesn't allow images, generated from their models to be sold on stock markets, but CGTrader does.
Really? I didn't know this is possible, I'm curious and I'll take a look how they can exactly restrict commercial rights. Thanks for the info
-
David Holz, midjourney CEO, said less than an hour ago that he doesn't like midjourney-generated images being put up for sale through stock agencies. He said that he is seriously considering banning the sale of midjourney-generated images through stock agencies.
I can't find any source for this.
And it would be a complete nonsense, if real: why in the earth Midjourney would donate this market to dozens of competitor?
Also from legal point of view it seems very hard to say "You have commercial rights BUT not for this specific market". David Holz cannot "ban" that market, at the contrary the market can "ban" midjourney, that has been already done by SS and Getty.
By the way... it's a strange world and anything it's possible :) I would like to read the original source with the words of David Holz
I have no information about the accuracy of the quote, but of course it would be no legal problem to restrict the usage in this way.
After all, stock agencies do it all the time. They sell commercial licences for the usage of images, but certain usages require an extended licence and some usages are forbidden even with an extended licence, notably reselling the image at other stock agencies.
-
I have no information about the accuracy of the quote, but of course it would be no legal problem to restrict the usage in this way.
After all, stock agencies do it all the time. They sell commercial licences for the usage of images, but certain usages require an extended licence and some usages are forbidden even with an extended licence, notably reselling the image at other stock agencies.
These are license restrictions for buyers.
Restriction of use for the copyright owner is completely different thing, much more difficult to manage. By the way, if someone do this, it means it's possible.
-
But Midjourney makes money selling access for commercial use, don‘t they?
So if artists can sell their designs to a commercial client, how will they stop the licensing via stock sites.
In the end they are just webshops.
If true this would mean midjourney cannot become a plug in for photoshop. They would be missing out on the biggest design market.
-
Easy. Microstock licenses are similar (at least standard licenses). Buyers can use images commercially for client's work, websites, publications etc. but cannot resell them. Same is with Turbosquid which someone mentioned before. So Midjourney can write something similar in their license. Why would they do or not do that, I don't want to assume right now, considering I haven't found that info anywhere either.
-
David Holz, midjourney CEO, said less than an hour ago that he doesn't like midjourney-generated images being put up for sale through stock agencies. He said that he is seriously considering banning the sale of midjourney-generated images through stock agencies.
Where?
Yesterday in a town hall meeting he organizes every Wednesday in discord from 12 noon to 4 pm. Pacific time
Very interesting meetings with David Holz, they also talked about the new /blend mode, with results almost of photography quality
Everyone can now use the /blend mode, it is amazing
-
Maybe he is upset because the stock agencies are coming after him to legally license the content.
He seems beyond naive when it comes to copyright.
If he scraped the entire internet and then charges people to use his ai he is a true hypocrit.
There is right way to do this.
But Adobe must be working 24/7 on their own ai and that will put him out of business in the end.
https://www.digitalcameraworld.com/news/midjourney-founder-basically-admits-to-copyright-breaching-and-artists-are-angry?fbclid=IwAR2v3PRf1VzQXNQDxwgnjwj6vz_EYpEv9v-pUxHYVdT-U4fcmE4JsafEZMw (https://www.digitalcameraworld.com/news/midjourney-founder-basically-admits-to-copyright-breaching-and-artists-are-angry?fbclid=IwAR2v3PRf1VzQXNQDxwgnjwj6vz_EYpEv9v-pUxHYVdT-U4fcmE4JsafEZMw)
-
@Lina
I doubt mj can restrict usage terms like don‘t sell on stock agencies if mj just stole it all from the internet. They hold no usage right to the training files, so I doubt they can restrict the sale of the remix.
Looks like a feast for Getty Lawyers.
-
Out of curiousity, if people are convertiing AI PNG images to illustrations , how are you doing this?
-
Maybe he is upset because the stock agencies are coming after him to legally license the content.
He seems beyond naive when it comes to copyright.
If he scraped the entire internet and then charges people to use his ai he is a true hypocrit.
There is right way to do this.
But Adobe must be working 24/7 on their own ai and that will put him out of business in the end.
https://www.digitalcameraworld.com/news/midjourney-founder-basically-admits-to-copyright-breaching-and-artists-are-angry?fbclid=IwAR2v3PRf1VzQXNQDxwgnjwj6vz_EYpEv9v-pUxHYVdT-U4fcmE4JsafEZMw (https://www.digitalcameraworld.com/news/midjourney-founder-basically-admits-to-copyright-breaching-and-artists-are-angry?fbclid=IwAR2v3PRf1VzQXNQDxwgnjwj6vz_EYpEv9v-pUxHYVdT-U4fcmE4JsafEZMw)
You and the author make a huge list of biased and leading assumptions. Without permission doesn't mean illegally and infringing. Her piece is angry and outraged and leads the readers by repeating her personal opinion as known facts. The headline itself "Midjourney founder basically admits to copyright breaching and artists are angry"? Does he actually admit copyright breach?
Without consent 5 times, "The fact that the Midjourney team "weren't picky" suggests that this data set has not been carefully curated and could be biased. ", cleverly spotted if your images have been used, is in fact The website searches the LAION-5B training data set, a library of 5.85 billion images, that is used to feed Stable Diffusion and Google’s Imagen. which isn't a secret at all.
"do you share the same rage as other artists and photographers on the subject?" ::) No I don't.
May 2022 - LAION-5B is released. https://www.infoq.com/news/2022/05/laion-5b-image-text-dataset/#:~:text=The%20Large-scale%20Artificial%20Intelligence%20Open%20Network%20%28LAION%29%20released,making%20it%20the%20largest%20freely%20available%20image-text%20dataset. (https://www.infoq.com/news/2022/05/laion-5b-image-text-dataset/#:~:text=The%20Large-scale%20Artificial%20Intelligence%20Open%20Network%20%28LAION%29%20released,making%20it%20the%20largest%20freely%20available%20image-text%20dataset.)
You can download the LAION AI set, with text and image matched results, for free. No one owns the right to the training files, they are public property.
I think you are right about SS and AS creating their own versions that will be reviewed and monitored and will attempt to be unbiased as well as culling out any accidentally included, infringing images. A more responsible approach.
-
Maybe he is upset because the stock agencies are coming after him to legally license the content.
He seems beyond naive when it comes to copyright.
If he scraped the entire internet and then charges people to use his ai he is a true hypocrit.
Yeah, same thing came to my mind, that he is upset at Getty coming after him, plus more users/buyers = more possible copyright problems, lawsuits and so on. And yes, definitely it is hypocritical, if he thinks it wasn't necessary to pay artists for using their images, then everything created with AI should also be free or at least available for personal use only.
@Lina
I doubt mj can restricting usage terms like don‘t sell on stock agencies if mj just stole it all from the internet. They hold no usage right to the training files, so I doubt they can restrict the sale of the remix.
Looks like a feast for Getty Lawyers.
Yeah, that might be possible.
-
@Lina
I doubt mj can restricting usage terms like don‘t sell on stock agencies if mj just stole it all from the internet. They hold no usage right to the training files, so I doubt they can restrict the sale of the remix.
...
1. you have no evidence they stole anything
2. you have no proof they lack usage rights other than your opinion
so your conclusion is illogical, but rather an ill-informed opinion,
-
Out of curiousity, if people are convertiing AI PNG images to illustrations , how are you doing this?
I use Gigapixel ai from topazlabs. Upsizes and converts to jpg. Works for me.
-
COBALT.
Thanks for the reply.
-
@Lina
I doubt mj can restricting usage terms like don‘t sell on stock agencies if mj just stole it all from the internet. They hold no usage right to the training files, so I doubt they can restrict the sale of the remix.
...
1. you have no evidence they stole anything
2. you have no proof they lack usage rights other than your opinion
so your conclusion is illogical, but rather an ill-informed opinion,
Did you read the interview???
Quote
„ There isn’t really a way to get a hundred million images and know where they’re coming from. It would be cool if images had metadata embedded in them about the copyright owner or something. But that's not a thing; there's not a registry. There’s no way to find a picture on the Internet, and then automatically trace it to an owner and then have any way of doing anything to authenticate it."
that is what he said.
If the ai was trained on perfectly legal content, he would have stated that by now, wouldn‘t he?
Would remove all doubt if mj files were legal to use for commercial purposes.
But everything in the interview makes it obvious he doesn‘t care about copyright at all and just scraped the internet.
You are the one suggesting he used legally suitable files, but Holz never mentions that. And it would have been an easy thing for him to do.
A feast for lawyers.
-
I really hope Adobe is working on their own AI. One that is even better than all that is out there now. Would keep them independent and they understand the media design business perfectly.
They can use the files we upload now as a bench mark and make sure their AI is superior.
-
I really hope Adobe is working on their own AI. One that is even better than all that is out there now. Would keep them independent and they understand the media design business perfectly.
They can use the files we upload now as a bench mark and make sure their AI is superior.
If the have their own AI, what do they need the AI "artist" contributors for?
-
David Holz, midjourney CEO, said less than an hour ago that he doesn't like midjourney-generated images being put up for sale through stock agencies. He said that he is seriously considering banning the sale of midjourney-generated images through stock agencies.
Where?
Yesterday in a town hall meeting he organizes every Wednesday in discord from 12 noon to 4 pm. Pacific time
Very interesting meetings with David Holz, they also talked about the new /blend mode, with results almost of photography quality
Everyone can now use the /blend mode, it is amazing
This person is talking garbage.
First he built a tool which disrupted the market and many people lost their job.
And now he is afraid that with stock agencies the market is getting disrupted.
-
David Holz, midjourney CEO, said less than an hour ago that he doesn't like midjourney-generated images being put up for sale through stock agencies. He said that he is seriously considering banning the sale of midjourney-generated images through stock agencies.
Where?
Yesterday in a town hall meeting he organizes every Wednesday in discord from 12 noon to 4 pm. Pacific time
Very interesting meetings with David Holz, they also talked about the new /blend mode, with results almost of photography quality
Everyone can now use the /blend mode, it is amazing
This person is talking garbage.
Have you seen the discord chat the person was talking about? Because I haven't and he apparently has. So why dismiss something as garbage if you weren't there to witness it?
And nowhere did he say David Holz was afraid that with stock agencies the market was getting disrupted. I doubt Holz is worrying about much other than his profit.
But stock agencies suing him because he used their images to train his AI without byuing licenses - That seems kind of a not too far fetched reason to consider banning the sale of midjourney images on stock agencies? Don't you think?
-
I really hope Adobe is working on their own AI. One that is even better than all that is out there now. Would keep them independent and they understand the media design business perfectly.
They can use the files we upload now as a bench mark and make sure their AI is superior.
If the have their own AI, what do they need the AI "artist" contributors for?
The ai is not a magic machine.
Somebody needs to create the useful content for the customers. That will be us.
Ai is a creative tool that designers and creators use in their workflow. It will probably have a Photoshop integration in some way.
The customers all have cameras and still buy our content. Some will use ai as well, but it will always be faster to just type in a keyword and then look at thousands of options to download, then to endlessly generate and tweak your prompts until you get what you need. there are so many possibilities for any theme.
-
The customers all have cameras and still buy our content.
That's because, contrarary to popular believe, creating a good photo is more than just pushing a button on a cheap camera. ::)
Unlike with midjourney, where everyone can enter a sentence.
-
A feast for lawyers.
That's for sure.
And now as an advocate for the other side. (opposing what I wrote yesterday)
Just because someone used LAION-5B which is free and public domain that doesn't mean the images in that 5 Billion images with data set are actually public domain.
In fact I could argue pretty easily that they are not. Not just because each one would have to be verified or that scraping is legal or illegal, or that artists styles are being copied or any of that.
The way they get these images with text is scraping for Alt-Text along with the images, then checking relevance, removing duplicates. But... someone entered that Alt-Text for the image. That means that if the website is copyrighted, which they are, then the text is also protected information, so the entire data set of LAION-5B is illegal and violating copyright laws.
The customers all have cameras and still buy our content.
That's because, contrarary to popular believe, creating a good photo is more than just pushing a button on a cheap camera. ::)
Unlike with midjourney, where everyone can enter a sentence.
Anyone CAN just push a button on a camera, same as anyone CAN just enter a sentence. There's more to entering a text prompt with the right requests and then there's editing after that as well.
AI is terribly flawed in so many ways that the people who want to demonize it, find obscure examples of good images and ignore hundreds of thousands of images that are flawed scrap. Faces, hands, body parts, machines, balance, physical logic, physics, and all kinds of other parts of AI images are flawed an impossible.
I can go to Midjourney (If I had access and an account) and enter the exact words and someone else and I won't get the same image. I won't get a finished work of art either. Just like I can own the same camera and lens and same settings and stand in the same place and... oh darn, I would get the same identical image, wouldn't I, just by pushing a button?
A feast for lawyers.
Yeah, worth repeating. That's who will decide all of this, then the courts and the appeals and after millions of dollars paid for that? Nothing we think or write here will change a speck of what will happen in the legal system in the future.
-
David Holz, midjourney CEO, said less than an hour ago that he doesn't like midjourney-generated images being put up for sale through stock agencies. He said that he is seriously considering banning the sale of midjourney-generated images through stock agencies.
Where?
Yesterday in a town hall meeting he organizes every Wednesday in discord from 12 noon to 4 pm. Pacific time
Very interesting meetings with David Holz, they also talked about the new /blend mode, with results almost of photography quality
Everyone can now use the /blend mode, it is amazing
This person is talking garbage.
First he built a tool which disrupted the market and many people lost their job.
And now he is afraid that with stock agencies the market is getting disrupted.
He said it in January 25th, 2023. I did not invented. In "office hours" town hall meeting David Holz said several times that he does not like mid journey images in stock portals and he is seriously considering to ban the sale of Midjourney generated content in stock photo web sites. I AM NOT LYING. He said it in front of more than 1000 mid journey members.
-
David Holz, midjourney CEO, said less than an hour ago that he doesn't like midjourney-generated images being put up for sale through stock agencies. He said that he is seriously considering banning the sale of midjourney-generated images through stock agencies.
Where?
Yesterday in a town hall meeting he organizes every Wednesday in discord from 12 noon to 4 pm. Pacific time
Very interesting meetings with David Holz, they also talked about the new /blend mode, with results almost of photography quality
Everyone can now use the /blend mode, it is amazing
This person is talking garbage.
First he built a tool which disrupted the market and many people lost their job.
And now he is afraid that with stock agencies the market is getting disrupted.
He said it in January 25th, 2023. I did not invented. In "office hours" town hall meeting David Holz said several times that he does not like mid journey images in stock portals and he is seriously considering to ban the sale of Midjourney generated content in stock photo web sites. I AM NOT LYING. He said it in front of more than 1000 mid journey members.
I am not pointing you. I am mentioning David Holz as an Idiot.
He is talking rubbish.
It funny to see that the AI makers are now afraid that with the use of AI the market is dirupting. As if they did a very generous job by inventing this.
I have people in my network who got fired by companies as their job is done by AI now. They are undergoing depression as their skills are taken over by AI.
I am not mentioning AI in images as particular in above case.
-
The customers all have cameras and still buy our content.
That's because, contrarary to popular believe, creating a good photo is more than just pushing a button on a cheap camera. ::)
Unlike with midjourney, where everyone can enter a sentence.
sure- but how many images have you sold that were created with 1 sentence and no additional work on your part?
Glendower: I can call the spirits from the vasty deep.
Hotspur: Why, so can I, or so can any man;
But will they come, when you do call for them?”
― William Shakespeare, King Henry IV, Part 1
-
@adh
I fully believe he said he might want to ban selling mj files on stock sites.
But the question is - if he has not licensed the training files for his ai and then for commercial use of the product, how can he legally stop anyone from uploading?
the pixels need a clear copyright track record.
And he is charging artists money for commercial use.
We had the example of 3d render companies that don't allow the reselling of the finished product, only for single client use.
But they hold the rights for their 3d models. Completely different situation.
the easiest thing would be is if mj license files for training purposes, the way dalle has done apparently, or other ais as getty claims have licensed from them for training.
then he can establish a track record that allows commercial use in a proper way.
but if he does not have the licensed legal rights for remixing pixels...then he is selling hot air. and mj might be liable themselves. In quite a big way.
We will see, the courts will settle this.
But since some companies have already properly licensed files, the technology can be used. there will be a legal protocol that all these ai companies follow in the end.
eta
if he is regularly online, why not ask him on the training files for his ai. Did mj legally license the files to remix pixels and offer fresh images for commercial use??
the way other companies have??
should be an easy question if he has his legal pathway in order.
gettyimages alone has over 700 million files, don't they? Together with a few other agencies there are certainly over 1 billion well keyworded, described, high quality images that can be licensed for commercial training of ais.
Very, very simple thing to do.
-
if he is regularly online, why not ask him on the training files for his ai. Did mj legally license the files to remix pixels and offer fresh images for commercial use??
the way other companies have??
should be an easy question if he has his legal pathway in order.
He has been asked this already and officially admitted that no, they did not legally license any of the images. He basically said he would have loved to compensate artists, but there just, very sadly, was absolutely no way for him to find out whom the images he used to train his AI belonged to.
Yep. He seriously claimed that. ::)
Here is the original quote, from an interview with David Holz with forbes:
No, there’s really no way to get a hundred million pictures and know where they’re coming from. It would be nice if the images had embedded metadata about the copyright owner or something. But it doesn’t matter; There is no registry. There’s no way to find a picture on the internet, and then automatically trace it to an owner and then do whatever it takes to authenticate it.
If only there was something like a huge image library with millions of images where each owner can be traced and compensated if his image was used. Oh, well, too bad there is not.... ::) There is really nothing David Holz could have done to compensate artists. Man, he really tried!
-
The training material. Here's where the files came from. https://laion.ai/blog/laion-5b/
Does anyone read or understand?
-
The training material. Here's where the files came from. https://laion.ai/blog/laion-5b/
Does anyone read or understand?
LOL - how can you possibly expect trolls to RYFM?
-
And where does it say that the datasets were created with files licensed from the copyright owners to be used for commercial purposes??
"To address this problem we present LAION 5B, a large-scale dataset for research purposes consisting of 5,85B CLIP-filtered image-text pairs"
It just says for "research purposes" and gives no explanation how the creators of the content were compensated or that they agreed to have their files used for datasets for training ai.
So they scraped the net and created datasets.
They explain their process in great detail, how they used a common crawl to scour the internet and downloaded files without asking anyone for permission, then refined the process to try and minimize files with visible watermarks.
But absolutely nowhere do they ask creators for licensing rights.
And did they exclude all the personal profile pics, family photos people post publicly on their social media pages? Or images of copyrighted products, designs and technology? Drawings of architectural plans??
Imagine downloading all music clips and sounds you can find on the internet, pairing it up with description text.
Obviously you can set up programs to do that.
But it doesn't make it legal.
"I found it on the internet so I can use it any way I like" .... it doesn't work that way.
Like I said, a feast for the lawyers.
They have zero rights to the files. And then they "distribute" the images they stole with a cc license...just makes it worse...they claim coypright on something they don't own.
It exactly matches the interview of the ceo of mj, who claimed it would be impossible to compensate creators or identify and properly license 100 million files.
Looks like these companies are walking DODOs, they committed the largest image theft in art history and are bragging about it on the internet...
-
Just look at how they talk about watermarks
"Watermarked images are a big problem when training generative models like DALL-E or GLIDE. To tackle this problem we trained a watermark detection model and used it to calculate confidence scores for every image in LAION-5B. Therefore we created a training dataset consisting of 90.000 images with 50% watermarked and 50% clean images. The majority of the watermarked images have been extracted from the LAION-400M KNN index through the use of several text prompts like “clip art watermark”, “cat watermark” or “landscape watermark”."
It is just a nuisance to them, no awareness at all of the legal reasons watermarks exist to protect files from theft.
And also no awareness that images without watermark are not free to steal and abuse, but are usually files professionally licensed, so that they can be used and shown without a watermark.
They could have easily licensed files for training, paid for it and gotten beautiful images without a watermark.
No, this is not funny at all.
ETA:
It is a solvable problem. they need to license files properly and when they make people pay to create files for commercial use, make sure the remixed pixels only come from licensed files.
-
They could have easily licensed files for training, paid for it and gotten beautiful images without a watermark.
...
It is a solvable problem. they need to license files properly and when they make people pay to create files for commercial use, make sure the remixed pixels only come from licensed files.
if it's so solvable why have none of the critics actually proposed how such a system might work?
- how do you find out who owns the rights to each image?
- how do you contact the owners, if any?
- how do you track owners' responses? ie, how do you create a database of artists? attaching a copyright doesn't usually include contact info, and few images even have that minimal information.
- how much should be paid to artists?
- how are payments calculated? per image at tiny fractions of pennies?
- how is payment made with knowing details such as paypal, bank acct or physical address? will a bank process checks for < a penny?
if you're going to complain and allege criminal liability you need to at least make a minimal effort to present a solution than can actually be discussed otherwise it's just more (redundant) hot air adding nothing to a conversation. and it is th e responsibility of the plaintiff to prove they have a case with hard evidence of wrongdoing
-
And where does it say that the datasets were created with files licensed from the copyright owners to be used for commercial purposes??
I just wrote the same a couple days ago. They used a public domain database, which was created by by matching alt-text tags with found images, and none of the images were licensed. Also I'd question using the alt-text description as web pages are copyrighted? Alternative text (alt text) is descriptive text which conveys the meaning and context of a visual item in a digital setting, such as on an app or web page. Just for anyone who doesn't know what that means
Looks like these companies are walking DODOs, they committed the largest image theft in art history and are bragging about it on the internet...
The courts will have to decide.
But there are different kinds of AI and the one that just uses data to train and the ai creates new images from what it has been trained, is creating new images. No infringing because they never used the original image. You can't sue someone for reading the dictionary, or a book, and using what they learn to write something new.
The software that makes collages and uses part of other images, is a different question. Is the new image a derivative, or based on fair use, because it's transformative? A bunch of lawyers are going to get rich on this.
-
They could have easily licensed files for training, paid for it and gotten beautiful images without a watermark.
...
It is a solvable problem. they need to license files properly and when they make people pay to create files for commercial use, make sure the remixed pixels only come from licensed files.
if it's so solvable why have none of the critics actually proposed how such a system might work?
- how do you find out who owns the rights to each image?
- how do you contact the owners, if any?
- how do you track owners' responses? ie, how do you create a database of artists? attaching a copyright doesn't usually include contact info, and few images even have that minimal information.
- how much should be paid to artists?
- how are payments calculated? per image at tiny fractions of pennies?
- how is payment made with knowing details such as paypal, bank acct or physical address? will a bank process checks for < a penny?
if you're going to complain and allege criminal liability you need to at least make a minimal effort to present a solution than can actually be discussed otherwise it's just more (redundant) hot air adding nothing to a conversation. and it is th e responsibility of the plaintiff to prove they have a case with hard evidence of wrongdoing
Uhm. The solution is: License images from microstock agencies.
Or are you trolling? Because it's hard to imagine you could not come up with that solution yourself. As a microstock contributor. In a microstock forum.
-
They could have easily licensed files for training, paid for it and gotten beautiful images without a watermark.
...
It is a solvable problem. they need to license files properly and when they make people pay to create files for commercial use, make sure the remixed pixels only come from licensed files.
if it's so solvable why have none of the critics actually proposed how such a system might work?
- how do you find out who owns the rights to each image?
- how do you contact the owners, if any?
- how do you track owners' responses? ie, how do you create a database of artists? attaching a copyright doesn't usually include contact info, and few images even have that minimal information.
- how much should be paid to artists?
- how are payments calculated? per image at tiny fractions of pennies?
- how is payment made with knowing details such as paypal, bank acct or physical address? will a bank process checks for < a penny?
if you're going to complain and allege criminal liability you need to at least make a minimal effort to present a solution than can actually be discussed otherwise it's just more (redundant) hot air adding nothing to a conversation. and it is th e responsibility of the plaintiff to prove they have a case with hard evidence of wrongdoing
Serious?
I've heard rumours that there are companies (called stock agencies) who are in the business of selling licenses of images. The seem to have solved the problem of identifying and paying respective image owners.
So how about starting the whole thing not by just scraping the internet to create an image database?
But yes, that would have cost money...
It's so much easier to just do the illegal (?) thing and after the fact just say "oh, it's so impossible to track down all those copyright owners, sorry..."
-
All Midjourney have to do is to add this text to their End User License Agreement:
"By downloading this image you agree not to sell/distribute it through stock agencies."
If the customer doesn't agree, he/she simply won't be able to download the generated image.
-
All Midjourney have to do is to add this text to their End User License Agreement:
"By downloading this image you agree not to sell/distribute it through stock agencies."
If the customer doesn't agree, he/she simply won't be able to download the generated image.
Sorry but this is not a solution.
First, it's simply a gift for dozens of other engines that give commercial use.
Second, it has already discussed, it's not so clear how to distinguish what you can and cannot do with your own images.
Someone gave examples with the restriction for the buyers who license content, but these are not relevant. These are restrictions for buyers, not for the copyright owner.
If it were simple as you said it would have already been done.
Force AI engine to pay the source IS the problem, not to lock the engine itself. AI is between us and obviously nothing will stop it to work
-
All Midjourney have to do is to add this text to their End User License Agreement:
"By downloading this image you agree not to sell/distribute it through stock agencies."
If the customer doesn't agree, he/she simply won't be able to download the generated image.
No copyright doesn't work that way.
they need explicit permission of the copyrightholders to start training their ais and remixing their pixels.
They cannot make any claims inlcuding limits on content they simply don't own.
It is just like in the music world. Look at how extremely expensive it can be if an artist even accidentally somewhere has a riff from a song of a well known band. They will get sued to hell and back.
"Oh, I heard it on the internet somewhere, so I just included it"..and I don't allow you to sell my remixed song commercially...
Does not work.
If you steal, you don't own.
End of story.
-
All Midjourney have to do is to add this text to their End User License Agreement:
"By downloading this image you agree not to sell/distribute it through stock agencies."
If the customer doesn't agree, he/she simply won't be able to download the generated image.
No copyright doesn't work that way.
they need explicit permission of the copyrightholders to start training their ais and remixing their pixels.
They cannot make any claims inlcuding limits on content they simply don't own.
It is just like in the music world. Look at how extremely expensive it can be if an artist even accidentally somewhere has a riff from a song of a well known band. They will get sued to hell and back.
"Oh, I heard it on the internet somewhere, so I just included it"..and I don't allow you to sell my remixed song commercially...
Does not work.
If you steal, you don't own.
End of story.
I'm not talking about copyright here, but getting access to their AI generator. They have the right to grant access to the people who promise not to upload the generated images on stock sites, and deny access to those who don't.
And as a customer you can't have copyright over images that you've never been allowed to generate in the first place, because you refused to agree to the company's terms.
(But if you have generated images in the past, they can't limit the way you can use those images retroactively, of course.)
-
...
The software that makes collages and uses part of other images, is a different question. Is the new image a derivative, or based on fair use, because it's transformative? A bunch of lawyers are going to get rich on this.
just to be clear, this is not the case for ANY of the AI we're talking about here; and such a use would be violation of both owner's copyright and agency TOS.
-
They could have easily licensed files for training, paid for it and gotten beautiful images without a watermark.
...
It is a solvable problem. they need to license files properly and when they make people pay to create files for commercial use, make sure the remixed pixels only come from licensed files.
if it's so solvable why have none of the critics actually proposed how such a system might work?
- how do you find out who owns the rights to each image?
- how do you contact the owners, if any?
- how do you track owners' responses? ie, how do you create a database of artists? attaching a copyright doesn't usually include contact info, and few images even have that minimal information.
- how much should be paid to artists?
- how are payments calculated? per image at tiny fractions of pennies?
- how is payment made with knowing details such as paypal, bank acct or physical address? will a bank process checks for < a penny?
if you're going to complain and allege criminal liability you need to at least make a minimal effort to present a solution than can actually be discussed otherwise it's just more (redundant) hot air adding nothing to a conversation. and it is th e responsibility of the plaintiff to prove they have a case with hard evidence of wrongdoing
Uhm. The solution is: License images from microstock agencies.
Or are you trolling? Because it's hard to imagine you could not come up with that solution yourself. As a microstock contributor. In a microstock forum.
ROFL! i don't need no stinkin' solutions because i never claimed there was a need for them! i accept the current situation re ML & creation of datasets. it's your cohort that demand payment - yet you haven't shown any way this could be done. it's content free - just 'sound & fury, signifying nothing'
i gave a list of serious questions that need to be addressed in order to meet your demands but you ignored all of them & just left insults
-
...
The software that makes collages and uses part of other images, is a different question. Is the new image a derivative, or based on fair use, because it's transformative? A bunch of lawyers are going to get rich on this.
just to be clear, this is not the case for ANY of the AI we're talking about here; and such a use would be violation of both owner's copyright and agency TOS.
And that's right. I don't know if any are using collage but the AI software that is modern and like DALL-E2, Canva/Stable Diffusion or Midjourney are text to image.
These create entirely new images from scratch, based on descriptive text. There's no infringing because the images are 100% new.
The training is based from an image, with the text and the AI is trained on style and features and properties, NOT by copying images.
Text to image is where entirely new images are created from scratch there is no infringing on existing images.
-
DALL-E could be use for Adobe Stock or another agency?
i only see this about.
https://platform.openai.com/docs/usage-policies/disallowed-usage (https://platform.openai.com/docs/usage-policies/disallowed-usage)
-
DALL-E could be use for Adobe Stock or another agency?
i only see this about.
https://platform.openai.com/docs/usage-policies/disallowed-usage (https://platform.openai.com/docs/usage-policies/disallowed-usage)
Not my field of expertise, but I guess it can get you, theoretically, in trouble.
You generate an AI image, submit it to Adobe Stock, where it finds it's way to a customer who uses it for political campaigning.
Sure, how the image is used is beyond your control as you cannot specify the usage conditions or context, but it's still an image that you generated via OpenAI and your responsibility to make sure it's usage is not in violation with the openAI policies?
-
DALL-E could be use for Adobe Stock or another agency?
i only see this about.
https://platform.openai.com/docs/usage-policies/disallowed-usage (https://platform.openai.com/docs/usage-policies/disallowed-usage)
Not my field of expertise, but I guess it can get you, theoretically, in trouble.
You generate an AI image, submit it to Adobe Stock, where it finds it's way to a customer who uses it for political campaigning.
Sure, how the image is used is beyond your control as you cannot specify the usage conditions or context, but it's still an image that you generated via OpenAI and your responsibility to make sure it's usage is not in violation with the openAI policies?
Same goes for this:
"Consumer-facing uses of our models in medical, financial, and legal industries; in news generation or news summarization; and where else warranted, must provide a disclaimer to users informing them that AI is being used and of its potential limitations"
Consumer-facing means dealing with people who buy products or services, so this includes a lot of commercial usage in medical, financial, legal, & news fields - That's quite a broad field of usage where DALL requires a note that the content is AI generated. But how do you make sure customers who buy the image from an agency add that note, when the agencies themselves do not require such a note from the customers, so the customers would not even know about this?
-
DALL-E could be use for Adobe Stock or another agency?
i only see this about.
https://platform.openai.com/docs/usage-policies/disallowed-usage (https://platform.openai.com/docs/usage-policies/disallowed-usage)
Not my field of expertise, but I guess it can get you, theoretically, in trouble.
You generate an AI image, submit it to Adobe Stock, where it finds it's way to a customer who uses it for political campaigning.
Sure, how the image is used is beyond your control as you cannot specify the usage conditions or context, but it's still an image that you generated via OpenAI and your responsibility to make sure it's usage is not in violation with the openAI policies?
I'd say everyone is correct, just one emphasis: "We don’t allow the use of our models for the following:"
No models, no problem?
-
lets face it we are witnessing the death of picture copyright (may be except editorial)
even today you can put a picture as a prompt and get back a similar but different picture copyright free, think what will happen in another couple of years
-
DALL-E could be use for Adobe Stock or another agency?
i only see this about.
https://platform.openai.com/docs/usage-policies/disallowed-usage (https://platform.openai.com/docs/usage-policies/disallowed-usage)
Not my field of expertise, but I guess it can get you, theoretically, in trouble.
You generate an AI image, submit it to Adobe Stock, where it finds it's way to a customer who uses it for political campaigning.
Sure, how the image is used is beyond your control as you cannot specify the usage conditions or context, but it's still an image that you generated via OpenAI and your responsibility to make sure it's usage is not in violation with the openAI policies?
I'd say everyone is correct, just one emphasis: "We don’t allow the use of our models for the following:"
No models, no problem?
Hm. Interesting. I did not interpret "models" as "people", I thought they refered to their AI APs as "models", because that's what they call them here:
https://platform.openai.com/docs/models/overview
And if that's what they mean by models then it's rather "no models, no DALL-E, no problem".
-
DALL-E could be use for Adobe Stock or another agency?
i only see this about.
https://platform.openai.com/docs/usage-policies/disallowed-usage (https://platform.openai.com/docs/usage-policies/disallowed-usage)
Not my field of expertise, but I guess it can get you, theoretically, in trouble.
You generate an AI image, submit it to Adobe Stock, where it finds it's way to a customer who uses it for political campaigning.
Sure, how the image is used is beyond your control as you cannot specify the usage conditions or context, but it's still an image that you generated via OpenAI and your responsibility to make sure it's usage is not in violation with the openAI policies?
I'd say everyone is correct, just one emphasis: "We don’t allow the use of our models for the following:"
No models, no problem?
Hm. Interesting. I did not interpret "models" as "people", I thought they refered to their AI APs as "models", because that's what they call them here:
https://platform.openai.com/docs/models/overview
And if that's what they mean by models then it's rather "no models, no DALL-E, no problem".
I'm not sure, but that's the way I read it. Mostly because of the parts like not for politics, not for medical and not for testimonial type of use. I was thinking images. But further reading, you're right. I'm wrong.
Yes in that section for text and code, it lists:
We don’t allow the use of our models for the following:
Illegal activity
Child Sexual Abuse Material or any content that exploits or harms children
Generation of hateful, harassing, or violent content
Generation of malware
Activity that has high risk of physical harm
Activity that has high risk of economic harm
Fraudulent or deceptive activity
Adult content, adult industries, and dating apps
Political campaigning or lobbying
Activity that violates people’s privacy
Engaging in the unauthorized practice of law, or offering tailored legal advice without a qualified person reviewing the information
Offering tailored financial advice without a qualified person reviewing the information
Telling someone that they have or do not have a certain health condition, or providing instructions on how to cure or treat a health condition
High risk government decision-making
While DALL-E says:
You are not allowed to use DALL-E to generate any of the following types of content:
Hate
Harassment
Violence
Self-harm
Sexual
Shocking
Illegal activity
Deception
Political
Public and personal health
Spam
So OK back to...
You generate an AI image, submit it to Adobe Stock, where it finds it's way to a customer who uses it for political campaigning.
Sure, how the image is used is beyond your control as you cannot specify the usage conditions or context, but it's still an image that you generated via OpenAI and your responsibility to make sure it's usage is not in violation with the openAI policies?
Open AI would have to object and go to the user and then the agency and then back to the innocent artist. I wonder if the terms of use on Adobe and SSTK, for example, also say you can't use them for the above proposes that Open-AI doesn't allow? What if the images aren't marked created by AI or created by Open AI / DALL-E? Very complicated.
-
Open AI would have to object and go to the user and then the agency and then back to the innocent artist. I wonder if the terms of use on Adobe and SSTK, for example, also say you can't use them for the above proposes that Open-AI doesn't allow? What if the images aren't marked created by AI or created by Open AI / DALL-E? Very complicated.
True, it's quite a chain a complaint would have to follow, but in the end, if it's a serious complaint, it will land on the contributor's desk. In most cases, chances are small anyone would really take the effort to sit it through I guess, but the problem with political campaigns for instance is that they have a high visibility and a potential debatable and polarizing exposure. We already had cases of political parties using editorial stock images pulled out of context for political campaigning. Media and independent journalists were very fast in identifying where the image came from and clarifying correct usage conditions.
Personally I'd like to steer away from any use of my content for certain use-cases like political campaigning for instance, but that option is not available when submitting content.
We cannot exclude certain context.
Again, not my field of expertise here, but to me, it sounds like the usage conditions for AI generated content by DALL-E or OpenAI are not in line with the usage conditions that stock agencies apply. Or they must have different usage conditions for AI generated content to their customers.
-
Open AI would have to object and go to the user and then the agency and then back to the innocent artist. I wonder if the terms of use on Adobe and SSTK, for example, also say you can't use them for the above proposes that Open-AI doesn't allow? What if the images aren't marked created by AI or created by Open AI / DALL-E? Very complicated.
True, it's quite a chain a complaint would have to follow, but in the end, if it's a serious complaint, it will land on the contributor's desk. In most cases, chances are small anyone would really take the effort to sit it through I guess, but the problem with political campaigns for instance is that they have a high visibility and a potential debatable and polarizing exposure. We already had cases of political parties using editorial stock images pulled out of context for political campaigning. Media and independent journalists were very fast in identifying where the image came from and clarifying correct usage conditions.
Personally I'd like to steer away from any use of my content for certain use-cases like political campaigning for instance, but that option is not available when submitting content.
We cannot exclude certain context.
Again, not my field of expertise here, but to me, it sounds like the usage conditions for AI generated content by DALL-E or OpenAI are not in line with the usage conditions that stock agencies apply. Or they must have different usage conditions for AI generated content to their customers.
BINGO! 🔔 🔔 🔔
We are not responsible for the license terms, nor are we allowed to add conditions like the restrictions that Open-AI and others have placed on use for images we have created using their system.
Telling the person who wants to license images, that there are restrictions, is the duty of the agency.
No the complaint will not come back to the contributor. First off the person using the image, and then the agency, and then a slim possibility that someone from the AI company would try to chase down an artist. They would go for deep pockets first, not after the little that we have.
-
We are not responsible for the license terms
That's a strange view of things. You think because you are not the one who made the license terms you bear no responsibility? Tell that to the next thief who buys your image and re-sells it on another agencie and watch him tell you that he did nothing wrong, because he "is not responsibility" for the agencie's terms that do not allow re-selling of images.
No, you are not responsible for the license terms, but you are responsible for following them if you want to use and sell DALL-E images. You agreed to these terms.
And if DALL-E says "This image must not be used in political/medical/news content" and you agreed to that, then you can't sell them to customers who use them in political/medical/news content and really claim that you bear no responsibility in this and there was nothing you could have done to prevent this?
-
We are not responsible for the license terms
That's a strange view of things. You think because you are not the one who made the license terms you bear no responsibility? ...
No, you are not responsible for the license terms, but you are responsible for following them if you want to use and sell DALL-E images. You agreed to these terms.
And if DALL-E says "This image must not be used in political/medical/news content" and you agreed to that, then you can't sell them to customers who use them in political/medical/news content and really claim that you bear no responsibility in this and there was nothing you could have done to prevent this?
but that's the point - if we do follow the guidelines & our images are sold to someone who violates the license terms, how are we responsible?
a bigger issue is there's no way to tell which AI engine was used
-
You're talking about something that has nothing to do with AI
It's called "sensitive use", and it's a well know issue, it's absolutely managed on agency side and has nothing to do with the content.
The "not allowed" list of DALL-E (and other engines) is made to prevent the creation of the content; at the contrary, the sensitive use of ANY content (AI generated or not) is an agency side problem
We are not responsible for the license terms
That's a strange view of things. You think because you are not the one who made the license terms you bear no responsibility?
Absolutely yes. The creator is resposible of content copyright, absolutely not of the terms of license given to the buyer.
Telling the person who wants to license images, that there are restrictions, is the duty of the agency.
This is absolutely correct, this is the truth, no way to say that's not the case!
My opinion, of course ;D
-
Open AI would have to object and go to the user and then the agency and then back to the innocent artist. I wonder if the terms of use on Adobe and SSTK, for example, also say you can't use them for the above proposes that Open-AI doesn't allow? What if the images aren't marked created by AI or created by Open AI / DALL-E? Very complicated.
True, it's quite a chain a complaint would have to follow, but in the end, if it's a serious complaint, it will land on the contributor's desk. In most cases, chances are small anyone would really take the effort to sit it through I guess, but the problem with political campaigns for instance is that they have a high visibility and a potential debatable and polarizing exposure. We already had cases of political parties using editorial stock images pulled out of context for political campaigning. Media and independent journalists were very fast in identifying where the image came from and clarifying correct usage conditions.
Personally I'd like to steer away from any use of my content for certain use-cases like political campaigning for instance, but that option is not available when submitting content.
We cannot exclude certain context.
Again, not my field of expertise here, but to me, it sounds like the usage conditions for AI generated content by DALL-E or OpenAI are not in line with the usage conditions that stock agencies apply. Or they must have different usage conditions for AI generated content to their customers.
BINGO! 🔔 🔔 🔔
We are not responsible for the license terms, nor are we allowed to add conditions like the restrictions that Open-AI and others have placed on use for images we have created using their system.
Telling the person who wants to license images, that there are restrictions, is the duty of the agency.
No the complaint will not come back to the contributor. First off the person using the image, and then the agency, and then a slim possibility that someone from the AI company would try to chase down an artist. They would go for deep pockets first, not after the little that we have.
I don't know Pete. This is what's written in Adobe Stock's Generative AI requirements:
You must have all the necessary rights to submit generative AI illustrations to Adobe Stock for licensing and use as described in our contributor terms (e.g., broad commercial use). You must review the terms of any generative AI tools that you use to confirm that this is the case before you submit any AI-generated content.
Do: Read the terms and conditions for generative AI tools that you use to ensure that you have the right to license all generative AI content that you submit to Adobe Stock under the contributor terms. For example, you cannot submit any content if you are not permitted to license it for commercial purposes.
Don’t: Use generative AI tools that are known or recognized as having serious flaws in their design or outputs (for example, tools which generate identifiable people or property from generic prompts).
Don’t: Submit works depicting real places, identifiable property (e.g., famous characters or logos), or notable people (whether photorealistic or - even caricatures)
So you must have the full rights to submit it for "broad commercial use".
I assume (again, not my fleld of expertise) that this also means: political campaigning or other contexts which are excluded from DALL-E or OpenAI's terms of conditions.
So I see that as a responsibility of the contributor: make sure you have the full rights before submitting.
Admittedly, agencies are on the lazy side here as they can easily identify AI generated content, and they can apply different customer terms on them if they would want to. And they can whitelist certain AI Generative Tools which terms are in line with the agencies terms.
It all sounds like a theoretical discussion and complaints in this area feel like a rather unlikely event to happen, but it's not impossible either. Remember Alex and his news stand images receiving a complaint via Alamy.
-
So you must have the full rights to submit it for "broad commercial use".
I assume (again, not my fleld of expertise) that this also means: political campaigning or other contexts which are excluded from DALL-E or OpenAI's terms of conditions.
So I see that as a responsibility of the contributor: make sure you have the full rights before submitting.
mmmm... I think there is still confusion between
the content
and
the use of the content
i this sentence :-)
-
You're talking about something that has nothing to do with AI
It's called "sensitive use", and it's a well know issue, it's absolutely managed on agency side and has nothing to do with the content.
The "not allowed" list of DALL-E (and other engines) is made to prevent the creation of the content; at the contrary, the sensitive use of ANY content (AI generated or not) is an agency side problem
We are not responsible for the license terms
That's a strange view of things. You think because you are not the one who made the license terms you bear no responsibility?
Absolutely yes. The creator is resposible of content copyright, absolutely not of the terms of license given to the buyer.
Telling the person who wants to license images, that there are restrictions, is the duty of the agency.
This is absolutely correct, this is the truth, no way to say that's not the case!
My opinion, of course ;D
Interesting points. I wonder how this is different than submitting an editorial image as commercial, or an image without model release, which the agency lets slip through? Who's responsible for the error? The contributor for having made an incorrect submission, or the agency for letting it slip through and offering it with wrong license terms?
I always thought that the contributor was still responsible in such cases.
-
Interesting points. I wonder how this is different than submitting an editorial image as commercial, or an image without model release, which the agency lets slip through? Who's responsible for the error? The contributor for having made an incorrect submission, or the agency for letting it slip through and offering it with wrong license terms?
I always thought that the contributor was still responsible in such cases.
Good question!
In this case I would say the author is resposable, because he does not own the copyright.
But you're still not considering the other side, the use of image:
Following your example, if I should put for commercial use something like, let's say, a well known and famous car...
The legal problem would eventually rise only and when the image would be used for commercial.
In other words, it's not the image that break the rules, it's the use of it that could do
-
Interesting points. I wonder how this is different than submitting an editorial image as commercial, or an image without model release, which the agency lets slip through? Who's responsible for the error? The contributor for having made an incorrect submission, or the agency for letting it slip through and offering it with wrong license terms?
I always thought that the contributor was still responsible in such cases.
Good question!
In this case I would say the author is resposable, because he does not own the copyright.
But you're still not considering the other side, the use of image:
Following your example, if I should put for commercial use something like, let's say, a well known and famous car...
The legal problem would eventually rise only and when the image would be used for commercial.
In other words, it's not the image that break the rules, it's the use of it that could do
Still trying to wrap my head around this.
I understand that there's a difference in having copyrights and terms of conditions.
Thanks for clarifying that.
I guess the latter, violating terms and conditions, would generally mean less trouble than violating copyright?
In the case of selling an AI Generated image that ends up in a political campaign... copyright is not the issue here, but terms and conditions might be? I mean: the contributor generated the image and made it possible to be used in a context that violates the terms and conditions of the AI Generative Tool?
-
Interesting points. I wonder how this is different than submitting an editorial image as commercial, or an image without model release, which the agency lets slip through? Who's responsible for the error? The contributor for having made an incorrect submission, or the agency for letting it slip through and offering it with wrong license terms?
I always thought that the contributor was still responsible in such cases.
Good question!
In this case I would say the author is resposable, because he does not own the copyright.
But you're still not considering the other side, the use of image:
Following your example, if I should put for commercial use something like, let's say, a well known and famous car...
The legal problem would eventually rise only and when the image would be used for commercial.
In other words, it's not the image that break the rules, it's the use of it that could do
Still trying to wrap my head around this.
I understand that there's a difference in having copyrights and terms of conditions.
Thanks for clarifying that.
I guess the latter, violating terms and conditions, would generally mean less trouble than violating copyright?
In the case of selling an AI Generated image that ends up in a political campaign... copyright is not the issue here, but terms and conditions might be? I mean: the contributor generated the image and made it possible to be used in a context that violates the terms and conditions of the AI Generative Tool?
If we upload something commercial, and we never know what use it could be used for in the future, then we can't upload anything AI because someone might use it for a political, medical or other disallowed use.
We have no control over the license or use restrictions?
-
then we can't upload anything AI because someone might use it for a political, medical or other disallowed use.
To be precise, you can't upload AI made with engine that have these rules. There are several that do not apply any restrictions!
We have no control over the license or use restrictions?
If I remember well, SS gives the ability to exclude sensitive use for images, for the full portfolio. I don't know if the option is still alive, and I don't remember if others give same choice
-
then we can't upload anything AI because someone might use it for a political, medical or other disallowed use.
To be precise, you can't upload AI made with engine that have these rules. There are several that do not apply any restrictions!
We have no control over the license or use restrictions?
If I remember well, SS gives the ability to exclude sensitive use for images, for the full portfolio. I don't know if the option is still alive, and I don't remember if others give same choice
That's indeed the point I was trying to make, or better, the thing I would like to see clarified, as I'm anything but an expert in the matter.
In this case: We can 't upload AI generated content to Adobe Stock if the terms of use of the AI engine (DALL-E) don't match the licensing conditions of Adobe Stock. Question mark.
Maybe some AI engines like stable diffusion or midjourney have lesser strict terms... I don't know.
Maybe some stock agencies offer the possibility to exclude the use of your content in certain context... options here are very limited I guess.
Anyhow, it seems like a slippery slope, and when I read the conditions and conditions of Adobe Stock regarding AI Generated content, they seem to put all the responsibility on the contributor. You have to own the copyright (as far as I understood still a legally unclear situation in many parts of the world) and make sure the content is suited for 'broader commercial' use (which includes some parts that are excluded by the terms and conditions of OpenAI/DALL-E).
-
Maybe some stock agencies offer the possibility to exclude the use of your content in certain context... options here are very limited I guess.
I know that SS has an option (or at least used to) to exclude sexual and political content, but that's it. I don't know of any agency that would allow you to exclude usage for news or medical content.
-
Maybe some stock agencies offer the possibility to exclude the use of your content in certain context... options here are very limited I guess.
I know that SS has an option (or at least used to) to exclude sexual and political content, but that's it. I don't know of any agency that would allow you to exclude usage for news or medical content.
I remember that too, not sure it's still an available option.
But if it is, it's useless as Shutterstock doesn't accept AI Generated content from their contributors.
Shutterstock will not allow AI-generated content to be submitted for sale on our platform. We want to ensure contributors can prove IP ownership of all submitted content and also want to be confident that artists are properly compensated if and when their work is used in AI training models. Given the availability of various AI content generation models in the marketplace, we are unable to verify the model source for most AI-generated content and therefore are unable to ensure all artists who were involved in the generation of each piece of content are compensated.
https://support.submit.shutterstock.com/s/article/Shutterstock-ai-and-Computer-Vision-Contributor-FAQ?language=en_US
-
Maybe some stock agencies offer the possibility to exclude the use of your content in certain context... options here are very limited I guess.
I know that SS has an option (or at least used to) to exclude sexual and political content, but that's it. I don't know of any agency that would allow you to exclude usage for news or medical content.
I remember that too, not sure it's still an available option.
But if it is, it's useless as Shutterstock doesn't accept AI Generated content from their contributors.
Shutterstock will not allow AI-generated content to be submitted for sale on our platform. We want to ensure contributors can prove IP ownership of all submitted content and also want to be confident that artists are properly compensated if and when their work is used in AI training models. Given the availability of various AI content generation models in the marketplace, we are unable to verify the model source for most AI-generated content and therefore are unable to ensure all artists who were involved in the generation of each piece of content are compensated.
https://support.submit.shutterstock.com/s/article/Shutterstock-ai-and-Computer-Vision-Contributor-FAQ?language=en_US
Yup, they don't accept AI content but you can create it and pay on their site, using the identical software?
Another version of that says:
"Why can’t I claim copyright to AI-generated content if I am the one who produced the query that generated the synthetic content?"
Because AI content generation models leverage the IP of many artists and their content, AI-generated content ownership cannot be assigned to an individual. Per our Contributor Terms of Service (Sections 13d and 13f), contributors must have proven IP ownership of all content that is submitted. As such, AI-generated content should not be submitted to Shutterstock.
-
While searching for my own Chile landscape images at Adobe, I found that the landscape section is also flooded with realistic looking AI generated images.
But they are usually still fantasy images, e.g. this one:
https://stock.adobe.com/de/search?filters%5Bcontent_type%3Aphoto%5D=1&filters%5Bcontent_type%3Aillustration%5D=1&filters%5Bcontent_type%3Azip_vector%5D=1&filters%5Bcontent_type%3Avideo%5D=1&filters%5Bcontent_type%3Atemplate%5D=1&filters%5Bcontent_type%3A3d%5D=1&filters%5Bcontent_type%3Aimage%5D=1&order=relevance&safe_search=1&limit=100&search_page=1&k=chile+atacama+desert&search_type=usertyped&acp=&aco=chile+atacama+desert&get_facets=0&asset_id=573097280
The mountain range is ok, but the dunes look completely different there. The AI dunes are more likely to come from the Arabian region.
This will be very confusing for the buyers at some point at Adobe. Also, not every buyer will know what AI generated images are. For me, it spontaneously looked like a real landscape at first.
-
A travel magazine won't use AI generated content.
But look at AI generated food images like muffins - great images.
And this is just the beginning.
https://stock.adobe.com/de/search?filters%5Bcontent_type%3Aphoto%5D=1&filters%5Bcontent_type%3Aillustration%5D=1&filters%5Bcontent_type%3Azip_vector%5D=1&filters%5Bcontent_type%3Avideo%5D=1&filters%5Bcontent_type%3Atemplate%5D=1&filters%5Bcontent_type%3A3d%5D=1&filters%5Bcontent_type%3Aimage%5D=1&k=muffin+ai&order=relevance&safe_search=1&limit=100&search_page=1&search_type=usertyped&acp=&aco=muffin+ai&get_facets=0&asset_id=573812777 (https://stock.adobe.com/de/search?filters%5Bcontent_type%3Aphoto%5D=1&filters%5Bcontent_type%3Aillustration%5D=1&filters%5Bcontent_type%3Azip_vector%5D=1&filters%5Bcontent_type%3Avideo%5D=1&filters%5Bcontent_type%3Atemplate%5D=1&filters%5Bcontent_type%3A3d%5D=1&filters%5Bcontent_type%3Aimage%5D=1&k=muffin+ai&order=relevance&safe_search=1&limit=100&search_page=1&search_type=usertyped&acp=&aco=muffin+ai&get_facets=0&asset_id=573812777)
-
A travel magazine won't use AI generated content.
But look at AI generated food images like muffins - great images.
And this is just the beginning.
https://stock.adobe.com/de/search?filters%5Bcontent_type%3Aphoto%5D=1&filters%5Bcontent_type%3Aillustration%5D=1&filters%5Bcontent_type%3Azip_vector%5D=1&filters%5Bcontent_type%3Avideo%5D=1&filters%5Bcontent_type%3Atemplate%5D=1&filters%5Bcontent_type%3A3d%5D=1&filters%5Bcontent_type%3Aimage%5D=1&k=muffin+ai&order=relevance&safe_search=1&limit=100&search_page=1&search_type=usertyped&acp=&aco=muffin+ai&get_facets=0&asset_id=573812777 (https://stock.adobe.com/de/search?filters%5Bcontent_type%3Aphoto%5D=1&filters%5Bcontent_type%3Aillustration%5D=1&filters%5Bcontent_type%3Azip_vector%5D=1&filters%5Bcontent_type%3Avideo%5D=1&filters%5Bcontent_type%3Atemplate%5D=1&filters%5Bcontent_type%3A3d%5D=1&filters%5Bcontent_type%3Aimage%5D=1&k=muffin+ai&order=relevance&safe_search=1&limit=100&search_page=1&search_type=usertyped&acp=&aco=muffin+ai&get_facets=0&asset_id=573812777)
You are right, the pictures are great. But the subject of food is something else. I can represent food imaginatively in reality as well.
But a certain real landscape looks like it looks.
For a general landscape description "desert dunes with mountain", the picture is perfectly ok.
But in my example, someone has put together a fantasy landscape and assigned it to a certain real place "Death Valley san pedro de atacama" - and that is simply wrong and misleading.
And if I as a travel magazine am looking for real images and have to click through a mountain of AI images, I would find that very annoying and maybe change the agency.
I think that Adobe, as the search results now turn out, will not do itself any favors in the long run.
-
A travel magazine won't use AI generated content.
But look at AI generated food images like muffins - great images.
And this is just the beginning.
https://stock.adobe.com/de/search?filters%5Bcontent_type%3Aphoto%5D=1&filters%5Bcontent_type%3Aillustration%5D=1&filters%5Bcontent_type%3Azip_vector%5D=1&filters%5Bcontent_type%3Avideo%5D=1&filters%5Bcontent_type%3Atemplate%5D=1&filters%5Bcontent_type%3A3d%5D=1&filters%5Bcontent_type%3Aimage%5D=1&k=muffin+ai&order=relevance&safe_search=1&limit=100&search_page=1&search_type=usertyped&acp=&aco=muffin+ai&get_facets=0&asset_id=573812777 (https://stock.adobe.com/de/search?filters%5Bcontent_type%3Aphoto%5D=1&filters%5Bcontent_type%3Aillustration%5D=1&filters%5Bcontent_type%3Azip_vector%5D=1&filters%5Bcontent_type%3Avideo%5D=1&filters%5Bcontent_type%3Atemplate%5D=1&filters%5Bcontent_type%3A3d%5D=1&filters%5Bcontent_type%3Aimage%5D=1&k=muffin+ai&order=relevance&safe_search=1&limit=100&search_page=1&search_type=usertyped&acp=&aco=muffin+ai&get_facets=0&asset_id=573812777)
You are right, the pictures are great. But the subject of food is something else.
For a landscape description desert dunes with mountain, the picture is perfectly ok.
But in my example, someone has put together a fantasy landscape and assigned it to a certain real place "Death Valley san pedro de atacama" - and that is simply wrong.
And if I as a travel magazine am looking for real images and have to click through a mountain of AI images, I would find that very annoying and maybe change the agency.
I think that Adobe, as the search results now turn out, will not do itself any favors in the long run.
4682 muffin renderings feels like a lot.. Soon it will be millions or billions. I think they'll need an exclude AI search button.
-
A travel magazine won't use AI generated content.
But look at AI generated food images like muffins - great images.
And this is just the beginning.
https://stock.adobe.com/de/search?filters%5Bcontent_type%3Aphoto%5D=1&filters%5Bcontent_type%3Aillustration%5D=1&filters%5Bcontent_type%3Azip_vector%5D=1&filters%5Bcontent_type%3Avideo%5D=1&filters%5Bcontent_type%3Atemplate%5D=1&filters%5Bcontent_type%3A3d%5D=1&filters%5Bcontent_type%3Aimage%5D=1&k=muffin+ai&order=relevance&safe_search=1&limit=100&search_page=1&search_type=usertyped&acp=&aco=muffin+ai&get_facets=0&asset_id=573812777 (https://stock.adobe.com/de/search?filters%5Bcontent_type%3Aphoto%5D=1&filters%5Bcontent_type%3Aillustration%5D=1&filters%5Bcontent_type%3Azip_vector%5D=1&filters%5Bcontent_type%3Avideo%5D=1&filters%5Bcontent_type%3Atemplate%5D=1&filters%5Bcontent_type%3A3d%5D=1&filters%5Bcontent_type%3Aimage%5D=1&k=muffin+ai&order=relevance&safe_search=1&limit=100&search_page=1&search_type=usertyped&acp=&aco=muffin+ai&get_facets=0&asset_id=573812777)
You are right, the pictures are great. But the subject of food is something else.
For a landscape description desert dunes with mountain, the picture is perfectly ok.
But in my example, someone has put together a fantasy landscape and assigned it to a certain real place "Death Valley san pedro de atacama" - and that is simply wrong.
And if I as a travel magazine am looking for real images and have to click through a mountain of AI images, I would find that very annoying and maybe change the agency.
I think that Adobe, as the search results now turn out, will not do itself any favors in the long run.
4682 muffin renderings feels like a lot.. Soon it will be millions or billions. I think they'll need an exclude AI search button.
The way the whole thing is developing, it would be better if buyers want AI images to actively press an include/optin button.
I was totally annoyed today during my search. One clicks beautiful images ... AI, AI, AI .... As a potential buyer, I'd be out of Adobe.
-
It will be hard for a microstock contributor to maintain income without AI images.
If a contributor don't do microstock full-time he will maybe work some hours a week for microstock.
Now there are working some computers for AI microstock images full-time.
Meaning lots more of new AI images uploads daily.
It don't have do be better images than your non AI images.
But your images gets lost in search with stagnating uploads and more overall AI uploads.
-
Do buyers want a useful image that meets their needs or do they want a real photo only? I know we care, but do they?
Oh no this is terrible, I was just getting ready to flood the market with the next big thing. ;)
(https://i.postimg.cc/vHC9mbBt/sliced-tomato-background-multi-colored-GB3-USM-web.jpg)
AI Created Sliced Tomatoes Backgrounds
-
Do buyers want a useful image that meets their needs or do they want a real photo only? I know we care, but do they?
Oh no this is terrible, I was just getting ready to flood the market with the next big thing. ;)
(https://i.postimg.cc/vHC9mbBt/sliced-tomato-background-multi-colored-GB3-USM-web.jpg)
AI Created Sliced Tomatoes Backgrounds
Pete, do we have to worry about you being so fixated on tomatoes? The tomato has long been called the love-apple (pomme d amour), and the fruit has been suspected of causing love mania ;)
I'm sure you're right.
There are a whole lot of subjects where buyers probably don't care how the image was created - food might be one of them, in my opinion. With landscapes, which are to be assigned to a real place, I would estimate it differently.
-
Do buyers want a useful image that meets their needs or do they want a real photo only? I know we care, but do they?
Oh no this is terrible, I was just getting ready to flood the market with the next big thing. ;)
AI Created Sliced Tomatoes Backgrounds
Pete, do we have to worry about you being so fixated on tomatoes? The tomato has long been called the love-apple (pomme d amour), and the fruit has been suspected of causing love mania ;)
I'm sure you're right.
There are a whole lot of subjects where buyers probably don't care how the image was created - food might be one of them, in my opinion. With landscapes, which are to be assigned to a real place, I would estimate it differently.
Obsessed with stale stock and why people keep making it? I really should have picked something better, like girl with a bicycle in a field, or lady jumping with a long piece of streaming silk, or maybe smiling person with a headset? Oh wait, diversity, multicultural, business handshake? ::)
You haven't recognized the real obsession is the triple cheeseburger? (no they don't sell, just for the people who monitor the form for inspiration)
True a landscape and especially if it hints of a real location, should be real. That's another area, away from AI where people will label things with keywords like "The Alps" when they are taken in South America. Just anything for a view and maybe another dime, by adding false words that are inaccurate. Then the buyer might think it's really the location and gets told later, they have shown something in error. Which makes them look bad.
I'd agree AI should be listed as AI when it's supposed to represent anything or anyplace or something that's supposed to be real. After that, backgrounds, sandy beach, mountains and a stream, generic, I'm not sure that's a problem.
Tomatoes are in the nightshade family and where back in history of Europe, thought to be poisonous. Wow as part Italian, I don't know what we'd do without tomatoes? Poison Apples!
But thanks to the Italians and pizza and people discovering that their pewter plates were the problem = metal poisoning, we're all happy and safe now. Now people think things like GMO plants, carbs, and Gluten are poison to our system. Those three have replaced salt, sugar, artificial sweeteners and food coloring.
Eggs were bad, now they have been cleared. Coffee and tea, come and go. Chocolate but the latest stir is dark chocolate which has been found to contain cadmium and lead. Wait, bottled water is neutered but still managed to be tainted by plastic. Free range chickens are a myth, mostly not free range and natural means they are allowed to eat any crap they find, aren't free of disease, and generally are no different from caged chickens, except being more expensive and smaller because of their poor health.
Wait a few more years, it will be something else.
Until then I'm growing tomatoes as food subjects, eggplant, some squash and gourds. The little herb garden, no wild guess needed, is Chives, Oregano, Basil and Cilantro. Oh and just so someone doesn't read this and wonder, my garden is 100% organic and has been forever. No poisons. The farthest I go is natural compost and manure.
Key: food is fun to photograph? Plop and shoot, then eat it! 8)
-
The whole AI thing is very sad for artists who have spent a lifetime honing their design, illustration and photography skills. Particularly since these AI programs learned to make their "art" by sampling the creative output of millions of genuine artists, living and dead. I certainly understand that time and technology will never stand still and many previous skilled professions have been rendered obsolete in the past, but this seems particularly devastating for artists, writers and other creatives.
AI also seems poised to eliminate many other mid-level, well-paying jobs such as in the tech industry, further eroding the middle class and increasing the wealth gap.
Welcome to the brave new world.
-
@UnclePete how did you predict this news? Mediterranean diet ‘associated with 23% lower risk of dementia’. But new fads come and go, there's a new one the Mono diet.
Cameras took away the art of creating images by hand. Acrylic paint isn't oil. Digital cameras took away years of learning and experience by real photographers.
-
@UnclePete how did you predict this news? Mediterranean diet ‘associated with 23% lower risk of dementia’. But new fads come and go, there's a new one the Mono diet.
Cameras took away the art of creating images by hand. Acrylic paint isn't oil. Digital cameras took away years of learning and experience by real photographers.
I consult using ancient secrets of the unknown, by holding a chain of good luck symbols in my teeth. The rabbit wasn't so lucky, it's foot was made into a key chain charm?
(https://i.postimg.cc/6pmkhL4W/Pete-Sorcerer-Gnome-Charms.jpg)
(or maybe not and this is just a joke image?)
Food fads, trends and new discoveries, especially old ones rediscovered, make up conversation and have for ages. People looking for a secret to better health but discovering some trick. Let me say "you can't fool Mother Nature" But when you start to read the latest trending better health discoveries, you will mostly find, Buy This Book, join this plan, nutritional supplements - which is as big or bigger industry, than the Big Pharma they criticize.
Then there are the outright nutty for sensation: "CAVEMAN DIET I eat raw beef, liver, kidneys and chicken – it’s transformed my life" Or one of the latest? The charcoal detox diet claims to help people lose weight by "detoxing" them. It involves periods of fasting and consumption of tea or juice drinks that contain charcoal.
A good healthful, balanced diet, watch your calories, get the nutrition without excess or starving your cells. Calories in is pretty simple, that's what makes people gain weight. If you don't use as much energy as you eat, you will start to store that excess. You don't need to run a marathon, just get some exercise, walk or swim.
I'm saving for a crystal ball. ;)
-
In the Adobe Stock Additional Terms it says:
7. Restrictions.
7.1. General Restrictions. You must not:
...
(E) use the Stock Assets in a manner, or in connection with a subject, that a reasonable person could consider unflattering, immoral, offensive, obscene, or controversial, taking into account the nature of the Stock Assets, examples of which could include ads for tobacco; adult entertainment clubs or similar venues or services; implied or stated endorsements of political parties or other opinion-based movements; or implying mental or physical impairment;
Soley FYI.
Interpret it as you may.
I'm not giving any legal opinions.
-
AS limited number of uploads (from Discord channel):
"As some of you may have noticed already, we have established some limits to the number of assets you can have pending moderation at once. If you reach that limit, you will have to wait until the moderation process is complete for your files before being able to submit new assets for consideration.
We hope this will unclog the moderation queue and result in better waiting times from now on.
The maximum number of assets you can have waiting for moderation will depend on several factors and will be adapted to the global queue.
Thanks for your understanding and patience."
-
Checked in on the Discord. Quote: "there were many people who sent 200, 300 AI images a day, I myself sent around 150 a day, then the moderators were overloaded" lol
-
Checked in on the Discord. Quote: "there were many people who sent 200, 300 AI images a day, I myself sent around 150 a day, then the moderators were overloaded" lol
Yeah, amazing, but usual talk there. Until someone gets blocked. :P
-
HI! Question about AI-generated images with people. Do I understand correctly that we can skip par of "Witness" in the Property/Model release if it is AI generated image, as there is no real person under 18, for whom witness would be actual? Please share your experience, how do you submit images with persons to Adobe. I don't have any person just to use as a signature giver as I see in lots of youtube videos as suggested.
-
HI! Question about AI-generated images with people. Do I understand correctly that we can skip par of "Witness" in the Property/Model release if it is AI generated image, as there is no real person under 18, for whom witness would be actual? Please share your experience, how do you submit images with persons to Adobe. I don't have any person just to use as a signature giver as I see in lots of youtube videos as suggested.
A property release is required for images that feature fictional people created with generative AI software. A witness is not required on that release.
More details on generative AI submission guidelines can be found here: https://helpx.adobe.com/stock/contributor/help/generative-ai-content.html
Thanks,
Mat Hayward
-
Generative Ai stock images should be separated to a different stock site, they should not mix with real photos and paintings and real human art.
Otherwise soon you will flood Adobe stock with millions of useless Ai images. Right now Ai looks cool cause it's something new. But all Ai is very similar and people uploading Ai too fast.
So soon this "cool Ai style" will be just "Ai similar garbage". I'm not against Ai, I'm just telling that for Ai and real photos there should be different places.
-
Was anyone able to upload images created by Midjourney to Adobe since they are not 4 megapixels?
-
Generative Ai stock images should be separated to a different stock site, they should not mix with real photos and paintings and real human art....
who determines what's real? or what's art?
no more useless than millions of uninspired shots from ai-assisted cameras from people who think they're artists because they can press abutton
if you can't compete in a changing world, dont blame the tools! stop whinging & make way for those who can
and buyers dont care how an image is made as long as it fits their needs
-
Was anyone able to upload images created by Midjourney to Adobe since they are not 4 megapixels?
have you read ANY of these threads? it's been answered many times..
-
Checked in on the Discord. Quote: "there were many people who sent 200, 300 AI images a day, I myself sent around 150 a day, then the moderators were overloaded" lol
I know a person in India who sent over 1000 images a day
-
Checked in on the Discord. Quote: "there were many people who sent 200, 300 AI images a day, I myself sent around 150 a day, then the moderators were overloaded" lol
I know a person in India who sent over 1000 images a day
Who can create and edit 1,000 images a day. That's some really dedicated work. I'd love to see what they are and how many sales?
-
Checked in on the Discord. Quote: "there were many people who sent 200, 300 AI images a day, I myself sent around 150 a day, then the moderators were overloaded" lol
I know a person in India who sent over 1000 images a day
Who can create and edit 1,000 images a day. That's some really dedicated work. I'd love to see what they are and how many sales?
Everyone, at least with AI.
All you need to do is enter a prompt and click the retry button over and over again. (You don't even have to do that manually, just use an auto clicker). 10 clicks per hour and you get 1000 different images for which you can even use the same keywords...
-
Checked in on the Discord. Quote: "there were many people who sent 200, 300 AI images a day, I myself sent around 150 a day, then the moderators were overloaded" lol
I know a person in India who sent over 1000 images a day
Who can create and edit 1,000 images a day. That's some really dedicated work. I'd love to see what they are and how many sales?
Everyone, at least with AI.
All you need to do is enter a prompt and click the retry button over and over again. (You don't even have to do that manually, just use an auto clicker). 10 clicks per hour and you get 1000 different images for which you can even use the same keywords...
Each one needs to be edited to remove the color bar in the lower right. Resizing could be done batch. Still, I haven't seen anything from Open AI that didn't need editing, and people who use the other, still need to edit for size and error's.
Yes some group could make 1,000 a day, but how many would be useful, or get downloads and how many are accepted? At some point if it's the same prompt, the images are going to be similar, and the keywords or description will be a big red flag.
Then there's the time to create, and to upload. 1000 minutes is 16 hours. The possibility that someone does 1,000 a day is highly unlikely, just from the simple math of time. 100 not so much of an issue.
If it was a team of people, not "someone", then 1,000 a day from a factory situation, that's also possible.
It's already being done:
African Studio = 1,305,871 stock photos, vectors, and illustrations are available royalty-free.
Lineartistpilot = 1,166,763 stock photos, vectors, and illustrations are available royalty-free.
-
Checked in on the Discord. Quote: "there were many people who sent 200, 300 AI images a day, I myself sent around 150 a day, then the moderators were overloaded" lol
I know a person in India who sent over 1000 images a day
Who can create and edit 1,000 images a day. That's some really dedicated work. I'd love to see what they are and how many sales?
Everyone, at least with AI.
All you need to do is enter a prompt and click the retry button over and over again. (You don't even have to do that manually, just use an auto clicker). 10 clicks per hour and you get 1000 different images for which you can even use the same keywords...
Each one needs to be edited to remove the color bar in the lower right. Resizing could be done batch. Still, I haven't seen anything from Open AI that didn't need editing, and people who use the other, still need to edit for size and error's.
Yes some group could make 1,000 a day, but how many would be useful, or get downloads and how many are accepted? At some point if it's the same prompt, the images are going to be similar, and the keywords or description will be a big red flag.
Then there's the time to create, and to upload. 1000 minutes is 16 hours. The possibility that someone does 1,000 a day is highly unlikely, just from the simple math of time. 100 not so much of an issue.
If it was a team of people, not "someone", then 1,000 a day from a factory situation, that's also possible.
It's already being done:
African Studio = 1,305,871 stock photos, vectors, and illustrations are available royalty-free.
Lineartistpilot = 1,166,763 stock photos, vectors, and illustrations are available royalty-free.
Thanks Pete for taking the subject of AI out of mythology and into reality mathematically.
In the end, real people have to do some work.
-
Each one needs to be edited to remove the color bar in the lower right. Resizing could be done batch. Still, I haven't seen anything from Open AI that didn't need editing, and people who use the other, still need to edit for size and error's.
Not sure what color bar you are talking about, my AI images don't have any.
Everyone who still claims AI images need some major editing is either lying to keep up the pretense that there is some actual human work involved or is simply using the wrong AI image generator.
Maybe this was true for the first gen Ai generators, but by now they produce perfect results at least 90% of the time.
-
Generative Ai stock images should be separated to a different stock site, they should not mix with real photos and paintings and real human art.
Otherwise soon you will flood Adobe stock with millions of useless Ai images. Right now Ai looks cool cause it's something new. But all Ai is very similar and people uploading Ai too fast.
So soon this "cool Ai style" will be just "Ai similar garbage". I'm not against Ai, I'm just telling that for Ai and real photos there should be different places.
All the. AI midjourney images are very similar to each other, it's as if several collaborators had "taken" among several the images of a specific designer.
-
Not sure what color bar you are talking about, my AI images don't have any.
Everyone who still claims AI images need some major editing is either lying to keep up the pretense that there is some actual human work involved or is simply using the wrong AI image generator.
Maybe this was true for the first gen Ai generators, but by now they produce perfect results at least 90% of the time.
Yes and if they don't you just reroll, no point wasting time editing when you can get a new version in a minute. Especially for stock. Getting images that are plenty good enough for most microstock buyers is very easy.
Also it is possible with a few simple promts to get very different style images. Any style you want really. If you just put in the most basic description then, yes, they can look similar.
-
Hi everyone,
Today, we are announcing our policy regarding generative AI content, and I’d like to share that we have begun accepting illustrations made using generative AI into our collection. We believe that generative AI tools can help our contributor community continue to create amazing content, and we believe in transparent, clear labeling for customers when it comes to this content.
We have prepared generative AI content submission guidelines and a page to answer common questions. We believe that our policy to accept AI generated content will enable contributors and customers to benefit from the value that AI generated content can bring.
In our Discord channel for Adobe Stock contributors, we opened a new channel #ai-generated-talk as a forum for addressing further questions. As always, I will also monitor this thread daily and will do my best to answer any questions not covered in the FAQ.
Submission guidelines: https://helpx.adobe.com/stock/contributor/help/generative-ai-content.html
FAQ: https://helpx.adobe.com/stock/contributor/help/generative-ai-faq.html
Discord community: https://discord.com/invite/adobestock
Thank you,
Mat Hayward
I never got my redemption code, contacted support multiple times and crickets. renewal tomorrow. what can i do?
-
How about sales, do AI images sell well?
-
Each one needs to be edited to remove the color bar in the lower right. Resizing could be done batch. Still, I haven't seen anything from Open AI that didn't need editing, and people who use the other, still need to edit for size and error's.
Not sure what color bar you are talking about, my AI images don't have any.
Everyone who still claims AI images need some major editing is either lying to keep up the pretense that there is some actual human work involved or is simply using the wrong AI image generator.
Maybe this was true for the first gen Ai generators, but by now they produce perfect results at least 90% of the time.
I clearly said Open AI, which you ignored and calling anyone a liar for the truth is kind of over the top? Yes there is actual human work involved, why do you keep denying that?
Please show me an AI image from whatever source you decide is the best. One of the perfect results, 90% of the time, you can pick the image.
Tell me how long it takes to create each image and show me one as an example, please, that didn't need any editing of any kind. You say, someone can just hit create, make 1,000 images and upload them. And they are all key worded and described easily because they are all the same request.
How much does it cost for each of these images to be created?
Time, how much time to make them, edit them, add metadata, upload?
How about sales, do AI images sell well?
I don't know. There are some that say because they are in the front at agencies, that they are popular, and the conclusion is, that means they are being sold. I'd like to see some data from people who are creating and uploading these? The kind that come from, type a command, get an image, add some words and description and poof, like magic, you made money. ;)
As asked, here's a Dall-E2 image that is not edited.
(https://i.postimg.cc/MK8hWqGZ/DALL-E-2023-01-03-09-30-17-a-tomato-that-is-a-clock.png)
Look in the lower right, color sample bar. It's 1024 x 1024. Other AI is probably better, but that's what's free.
Edge / Bing unedited
(https://i.postimg.cc/652V1DBj/mountain-climbers-into-the-mist-snow-covered-ridge.jpg)
Also 1024 and has a watermark
A person a blob and a dog?
Make that 4MP and here's a crop
(https://i.postimg.cc/bJ5DxwSz/mountain-climbers-into-the-mist-snow-covered-ridge-200-percent-crop.jpg)
That won't pass anywhere.
1,000 a day from someone? Doesn't that strike anyone as possibly a bit exaggerated to make the demon AI more evil?
Yes, I'm using the wrong AI generator because I'm not really a fan of any of them. If I could create 100 new images a day, using AI, that are marketable and will get downloads and make a profit, I might change my mind. I don't think high volume AI images are the answer to making a profit. I might be wrong?
-
Just my experience on Adobe, so far 278 accepted gen ai files, 31 downloads.
But I also spent 1000 euros on credits for various ais for commercial use.
I started uploading in early December, so maybe for such a short time it is not that bad. Also I don't have extremely generic files that would sell in high volume.
I am trying to do something different. Not using Midjourney, so hopefully I will have a collection that looks a little different to what others offer.
But for the money, perhaps doing it all Midjourney style is best, I don't know. It is probably the look the buyers want until they get tired of it. But that might take several years, it is like the instagram filter craze.
-
Just my experience on Adobe, so far 278 accepted gen ai files, 31 downloads.
But I also spent 1000 euros on credits for various ais for commercial use.
I started uploading in early December, so maybe for such a short time it is not that bad. Also I don't have extremely generic files that would sell in high volume.
I am trying to do something different. Not using Midjourney, so hopefully I will have a collection that looks a little different to what others offer.
But for the money, perhaps doing it all Midjourney style is best, I don't know. It is probably the look the buyers want until they get tired of it. But that might take several years, it is like the instagram filter craze.
That "look" will probably also evolve, broaden, and change, if not with Midjourney or DALL-E then with competitors, and maybe even much faster as we think, considered the speed of the development of AI we have seen so far. People will also develop more sophisticated prompting skills (as many already do now) and on the other hand regulations and legal restrictions or boundaries will be set by governments regarding the operating and use of AI.
I think it's fair to say that nobody exactly knows which direction AI will go, how big the impact exactly will be, and in which position it will find a mature state of existence.
We can only make educated guesses.
I also think traditional stock agencies will find themselves in an existential crisis at some point: will they remain marketplaces for content, or will they evolve in brokers for AI-training datasets, or will they become companies that offer AI engines to generate content. Currently some of the agencies are all of that, and I wonder whether such a broad business model will remain manageable.
Regarding dataset training: I see Shutterstock accepting nearly everything I feed them recently, almost no rejections at all anymore, and this is definitely a different strategy then a few months ago. Seems they are very hungry for content at the moment. At the same time, I've seen Adobe becoming more restrictive, with more rejections. But maybe that's just my anecdotal experience.
-
Shutterstock takes gen ai content?
I see both the invested money and time as a way to learn a new technology. Gradually I will learn what buyers like to buy from me.
Also went a little wild with spaceships, robot and fantasy art..don't think there is a big market for that :)
-
Each one needs to be edited to remove the color bar in the lower right. Resizing could be done batch. Still, I haven't seen anything from Open AI that didn't need editing, and people who use the other, still need to edit for size and error's.
Not sure what color bar you are talking about, my AI images don't have any.
Everyone who still claims AI images need some major editing is either lying to keep up the pretense that there is some actual human work involved or is simply using the wrong AI image generator.
Maybe this was true for the first gen Ai generators, but by now they produce perfect results at least 90% of the time.
I clearly said Open AI, which you ignored and calling anyone a liar for the truth is kind of over the top? Yes there is actual human work involved, why do you keep denying that?
Please show me an AI image from whatever source you decide is the best. One of the perfect results, 90% of the time, you can pick the image.
Tell me how long it takes to create each image and show me one as an example, please, that didn't need any editing of any kind. You say, someone can just hit create, make 1,000 images and upload them. And they are all key worded and described easily because they are all the same request.
How much does it cost for each of these images to be created?
Time, how much time to make them, edit them, add metadata, upload?
How about sales, do AI images sell well?
I don't know. There are some that say because they are in the front at agencies, that they are popular, and the conclusion is, that means they are being sold. I'd like to see some data from people who are creating and uploading these? The kind that come from, type a command, get an image, add some words and description and poof, like magic, you made money. ;)
As asked, here's a Dall-E2 image that is not edited.
(https://i.postimg.cc/MK8hWqGZ/DALL-E-2023-01-03-09-30-17-a-tomato-that-is-a-clock.png)
Look in the lower right, color sample bar. It's 1024 x 1024. Other AI is probably better, but that's what's free.
Edge / Bing unedited
(https://i.postimg.cc/652V1DBj/mountain-climbers-into-the-mist-snow-covered-ridge.jpg)
Also 1024 and has a watermark
A person a blob and a dog?
Make that 4MP and here's a crop
(https://i.postimg.cc/bJ5DxwSz/mountain-climbers-into-the-mist-snow-covered-ridge-200-percent-crop.jpg)
That won't pass anywhere.
I wanted to recreate your images to show you what can be done in seconds, but I am not sure what you were trying to go for in the first image. Tomatoe clock?
(https://i.postimg.cc/d0Z0B3kM/Kahvi-highly-luminous-detailed-image-of-a-red-tomatoe-shaped-al-dc273076-15b6-4aea-b4a8-811a6b760d62.png)
(https://i.postimg.cc/ncTMV6bW/Kahvi-highly-luminous-detailed-image-of-snow-covered-mountain-w-9735f6ca-961b-4b2e-8b6b-3e6e17c85cbe.png)
I think DALL-E currently is lightyears behind every other AI image generator. I just looked at the curated AI gallery on Shutterstock today and in my opinion all images there looked really bad.
-
Shutterstock takes gen ai content?
I see both the invested money and time as a way to learn a new technology. Gradually I will learn what buyers like to buy from me.
Also went a little wild with spaceships, robot and fantasy art..don't think there is a big market for that :)
Oh, sorry for the confusion, I mean just stills. I don't upload AI generated content so far.
I just notice that Shutterstock's rejection policy seems to be way less strict than it was a few months ago.
Spaceships and space exploration is a very actual topic nowadays, so I think it's very much worth it to experiment with it.
It's also a topic that suits AI very much, as not many stock photographers travel the universe :)
-
The Ai-generated images sell pretty well, but not by themselves.
My best seller has been download at Adobe 52 times - but as all my AI images it went through a careful postprocessing in Photoshop and other dedicated software. Not only to remove the obvious errors (too many fingers, missing eyes etc.) but also to give them a particular light and style - they must reflect the way I feel, not MidJourney.
Also devising the right prompt is no kid's play: the AI is still too primitive to guess accurately enough what one has in mind upon a few words.
Last but definitely not least, the human creativity lies in the idea behind the prompt - mediocre prompt, mediocre outcomes. I've been waiting long for someone to mention this here, but so far in vain...
-
Spaceships and space exploration is a very actual topic nowadays, so I think it's very much worth it to experiment with it.
It's also a topic that suits AI very much, as not many stock photographers travel the universe :)
True. We will soon get space hotels and space tourism.
Must make sure to cover that niche early...
-
The Ai-generated images sell pretty well, but not by themselves.
My best seller has been download at Adobe 52 times - but as all my AI images it went through a careful postprocessing in Photoshop and other dedicated software. Not only to remove the obvious errors (too many fingers, missing eyes etc.) but also to give them a particular light and style - they must reflect the way I feel, not MidJourney.
Also devising the right prompt is no kid's play: the AI is still too primitive to guess accurately enough what one has in mind upon a few words.
Last but definitely not least, the human creativity lies in the idea behind the prompt - mediocre prompt, mediocre outcomes. I've been waiting long for someone to mention this here, but so far in vain...
I don't see anything special looking at the prompts on the Midjourney community showcase site, unlike the results. Not to mention that many of the prompts include artists' names.
-
Each one needs to be edited to remove the color bar in the lower right. Resizing could be done batch. Still, I haven't seen anything from Open AI that didn't need editing, and people who use the other, still need to edit for size and error's.
Not sure what color bar you are talking about, my AI images don't have any.
Everyone who still claims AI images need some major editing is either lying to keep up the pretense that there is some actual human work involved or is simply using the wrong AI image generator.
Maybe this was true for the first gen Ai generators, but by now they produce perfect results at least 90% of the time.
Perfect results, you're joking.
-
Each one needs to be edited to remove the color bar in the lower right. Resizing could be done batch. Still, I haven't seen anything from Open AI that didn't need editing, and people who use the other, still need to edit for size and error's.
Not sure what color bar you are talking about, my AI images don't have any.
Everyone who still claims AI images need some major editing is either lying to keep up the pretense that there is some actual human work involved or is simply using the wrong AI image generator.
Maybe this was true for the first gen Ai generators, but by now they produce perfect results at least 90% of the time.
Perfect results, you're joking.
Nope.
-
I see both the invested money
What are you investing in? Isn't it free to create images in artificial intelligence?
-
The free section of midjourney does not include commercial use. The paid version does.
-
The free section of midjourney does not include commercial use. The paid version does.
And how much do you have to pay for the paid version?
-
Each one needs to be edited to remove the color bar in the lower right. Resizing could be done batch. Still, I haven't seen anything from Open AI that didn't need editing, and people who use the other, still need to edit for size and error's.
Not sure what color bar you are talking about, my AI images don't have any.
Everyone who still claims AI images need some major editing is either lying to keep up the pretense that there is some actual human work involved or is simply using the wrong AI image generator.
Maybe this was true for the first gen Ai generators, but by now they produce perfect results at least 90% of the time.
Perfect results, you're joking.
Artist wins and rejects the prize.
https://www.msn.com/en-xl/news/other/artist-refuses-photography-award-after-revealing-his-picture-was-ai-generated/ar-AA19Yj8M (https://www.msn.com/en-xl/news/other/artist-refuses-photography-award-after-revealing-his-picture-was-ai-generated/ar-AA19Yj8M)
For me, working with AI image generators is a co-creation, in which I am the director. It is not about pressing a button—and done it is.” The images are "fake memories of a past, that never existed, that no-one photographed," created by putting them through AI image generators between 20 and 40 times, Eldagsen says on his website.
-
I've tried midjourney, DALL-E & Adobe Firefly Beta (which is personal use only). Spent $15 in credits on Open AI (DALL-E) and the results still needed a lot of post-processing. Haven't used midjourney in months so not sure if it's gotten better but I had the best results with Adobe Firefly. Since the non-commercial beta is free I also spent several hours straight working on a concept and then used various pieces to make a composite as well as tweaking others but what I had to work with was better than I'd gotten with the other two. I even managed to get a couple cute fantasy concepts that didn't need post-processing.
Based on this experience, I can't see how anyone can generate 1000 images for commercial use on any of these platforms without spending a fair amount of $$$ (far more than what I've spent) and a lot of time. It's quicker than drawing for sure, and quicker than photography too, but the results are much more unpredictable and the results look more similar than two photos of the same place.
(I once took a 9-day trip with a friend to various European countries to shoot stock. She and I spent much of our time together traveling to the exact same places, and our shooting styles aren't vastly different. Yet, out of many thousands of images, we didn't take any that looked the same. Yet, with AI, especially with midjourney, but with the others too, it is easy to spot generative AI - results are much more similar than when two people "press a button," but I'm sure that as the technology and user's skill (I agree there is skill involved), this may change.
It requires a different kind of creativity akin to what is required to come up with stock photo concepts, so you are more of an art director, rather than an artist, though you are that too when you post-process the images, or even as you decide which images to iterate.
It's like using the Adobe "artistic" Neural filters, fun and creative, but after a while, it just does the same thing and gets rather boring.
I don't think AI will take our jobs. People are still painting in realistic styles, despite 100 years of photography. But maybe that's wishful thinking.
-
The free section of midjourney does not include commercial use. The paid version does.
And how much do you have to pay for the paid version?
https://docs.midjourney.com/docs/plans
-
The free section of midjourney does not include commercial use. The paid version does.
And how much do you have to pay for the paid version?
https://docs.midjourney.com/docs/plans
Did they end the free plan or just a temporary halt? I never signed up.
-
The free section of midjourney does not include commercial use. The paid version does.
And how much do you have to pay for the paid version?
https://docs.midjourney.com/docs/plans
Did they end the free plan or just a temporary halt? I never signed up.
Don't know. I'm waiting for Firefly's inclusion in Photoshop.
-
The free section of midjourney does not include commercial use. The paid version does.
And how much do you have to pay for the paid version?
https://docs.midjourney.com/docs/plans
I think it’s easier, faster and cheaper for a buyer to buy everything he needs on the stock than to pay and then edit it all. But for stockers, this tool can be very useful in order to sell it later.
-
The free section of midjourney does not include commercial use. The paid version does.
And how much do you have to pay for the paid version?
https://docs.midjourney.com/docs/plans
Did they end the free plan or just a temporary halt? I never signed up.
Don't know. I'm waiting for Firefly's inclusion in Photoshop.
I forgot about that. There sure is some interesting competition going on. And then there will be the usual, sorting out as the market is established for the top versions, and the rest shut down. If that happens and I have CC for two more years, I'll have to hope it's better than Open AI / DALL-E.
Probably named in honor of Rufus T. Firefly, who was the president of Freedonia.
(https://i.postimg.cc/rw25hTq8/Duck-Soup.gif)
-
firefly is also a cult scifi series
https://en.wikipedia.org/wiki/Firefly_(TV_series)
-
I keep looking at the new uploads for gen ai content.
While the content is often stunningly beautiful, a lot of the content is very, very similar. I suppose this is because everybody is copying everyone and mostly it all comes from midjourney.
So...I am wondering, has Adobe considered checking any incoming content against existing content and to give the content that is really something new a boost in the algorithms?
This would of course also work for normal fotos, millions of files of plates of spaghetti on red and white napkin...if someone uploads a plate on a pink and blue napkin, it would be something different for that theme and would deserve an intial boost.
Obviously then the customers will then effect the real positions by lightboxing and buying or ignoring these files.
Adobe could use ai recognition in the database to diversify the results.
Perhaps Adobe is already doing this, but gen ai is bringing "similar" (copycat) content to new levels, I was wondering if something like this would help.
Adobe could even implement a "similar" checker or tester for contributors. We could have an application that allows us to test for "visual uniqueness" of content.
"Only 0.3 percent of files are visually similar for these keywords"...or something like that.
Or you upload a series and the software marks the files that are more unique.
You then still need to decide if the more unique file is a useful file, but having some visual feedback could give interesting guidance and help to diversify the overall collection.
Also would be useful for all the newbies, who just sort by downloads and then copy the highest ranking file.
Just a little idea.