You can check if a file was created using gen ai tools in this link.

https://verify.contentauthenticity.org/inspect

All I see in the description is that they use "history and identity data attached to images". If there is nothing attached, they can't check anything. I just tried with both one of my real photos (that actually had Exif and metadata attached!) and then with an AI image. Results for both images: "No results found".

So much about that....


I doubt there will ever be a relyable tool for detecting AI. Open Source had developed a tool to detect AI text. They abandoned the project 1 or 2 months ago, because it only had a sucess rate of like 40%. That's worse than guessing.

I do wonder however, if the "opted out people" are able to resubmit their images if they have been marked "eligible for data licensing".   

No, they can't. That's the problem and why I am so upset about it. You get the message that the image was already accepted if you try to resubmit.
So, if the reviewer made a mistake - you cannot resubmit. And even IF the reviewer did not make a mistake - almost every problem, from noise, to lens dust to exposure and even focus can be fixed nowadays. But Shutterstock now won't let you fix a problem and resubmit anymore. Once the image has been accepted for data licensing, it is basically "lost" to you forever.

Quote from: Uncle Pete on September 15, 2023, 11:29

Quote from: RalfLiebhold on September 15, 2023, 08:26

Quote from: Her Ugliness on September 13, 2023, 10:07

Worst of all is that you now cannot access the image and delete it. Today, for the first time in months or maybe even a year, I had an image rejected and I know it is a perfectly technical fine image with commercial value. So I want to re-submitt it, but I cannot delete the old one that went to "data licensing". That's really upsetting.

Yes, you can view and delete all these " Eligible for data licensing" photos .  Just go to the reviewed content, in here you can click on "data catalog",which you can find at the affected image.

And is there a delete button now? There wasn't before. When some of us wrote SS answered, these would be archived, in case we changed our mind about data licensing. HA! More like held hostage and what right do they have to archive our rejected submissions? Plus, lets say I want to make it exclusive elsewhere? How do I do that when SS has a copy on file.

They did say they would review the process and our ability to delete our own images. Did they?

Pete, like I said you can delete them now.

Where is that delete button? I do not see this in my reviewed tab and the image does't show up anywhere else. I also do not have any link saying "data catalog" to click on. Where do you see this?
 There are only the three dots and when I click on them the only option is "See large preview"

And I only talked with Shutterstock support yesterday about this and was specifically told that the image accepted for data licensing cannot be deleted by me and they refuse to delete it. 

istock put up a notice/reminder that they do not accept images using any generative ai tools including generative fill etc...

Sadly most AI contributors will not care and istock will not be able to tell. I think, at least for now, if you look at an AI image in microstock-suitable enhanced full size, in 90% of all cases you can tell it's AI generated and not a real photo - at least for AI images that are supposed to look like photos. With illustrations it is much more difficult.
But if it is actually a real photo and just a part of it was altered with AI - How is a reviewer supposed to tell whether it was altered with traditional photoshop methods or with Firefly?

But I still appreciate the sentiment. iStock certainly was not one of my favorite agencies due to their extremely small 1cent commissions, but they have gained back some of my respect with their anti-AI stance and lawsuit.

From a customer's point of view the only time I did this was with illustrations when I need multiple illustration in the same artistic style for the same project.
For photos I do not really see the point, at least for myself.
I think customers do it from time to time. I can even tell when a lot of images from the same topic are being sold at the same time/to the same location. But this only happened with some topics that I am heavily focusing on in my port. If you have a lot of photos from, for example, a certain location and a customer wants to write an article about it, then it doesn't seem too far fetched to assume that he or she might browse your port for more if he found one photo he/she likes through the search.
But I don't think a pohoto editor who is searching for a single photo of, for example, a smoking elderly woman, finds one and then randomly browses a contributor's port when he is not looking for any more images of that topic.

 I really do not think any microstock agency has any real interest in cretaing a really "honest" or "fair" or "ethical" compensation system for AI training. They are just throwing these words around because they hope it will be a selling point for their AI product to customers, that's all. 

Also, I don't think you can pay contributors per individual image that was used to create an AI image, because this is not how it works. The AI does not for example  "use 100 contributor files to create an image so in their neural network". The AI uses ALL images for it's training and then it does not use individual images once it creates an AI image. It uses it's 'knowledge' of all images it has ever 'seen".
 So basically all 360+ Million images in Adobe's database are used whenever Firefly creates AI content. And even IF you wanted to compensate these contributors for each individual use (which Adobe has no interest in. They want to earn money through the usage of an AI and "replace" having to pay real humans after all. Otherwise there would be no profit in it for Adobe if they still had to pay us), so if you say each Ai image is "worth" $0.10 that would mean Adobe would have to divide this amount through the over 360 Million images in their database and that would come down to an amount that is so small that it would not even accumulate to a single cent in 100 years....
 And here you can get a pretty good idea of how "worthless" our images have become thanks to AI.

Quote from: Her Ugliness on September 13, 2023, 08:47

Mat, do I understand this correctly, that the payment is solely based on portfolio sice and sales, not on whether the images were actully used to train your AI? Because if that's the case  that basically means that someone who has not done a single real photo or illustration in all his life, but has only created tons of AI content based off other people's hard work also got the "compensation". So they basically got PAID for generating AI images based off the work of people who never got paid for it.
That's a far cry from Adobe's statement that you were "developing generative AI responsibly, with creators at the center."

The initial bonus is based on the all-time total number of approved images you submitted to Adobe Stock and the number of licenses that those images generated in the 12-month period between June 3rd, 2022 to June 2nd, 2023. The bonus is weighted towards license.

-Mat Hayward

Yes, I can read. I understood that very well. I just wanted to clarify whether that REALLY means that you also compensated people who only submit AI images, so never had any images to offer for your training in the first place. Because that seemed insane. But apparently that's the case as you just copied the info from the mail Adobe sent out without adressing the actual issue.
Very "fair". 
So far, of all the agencies that use their database to create AI content, Adobe seems actually to be the most unethical one! No opt-out option AND giving money that was meant as compensation for having our real photos and illustrations used to train your AI to people who only use our content to generate AI images. I am sure these tons of new "contributors" that only submit AI content are overjoyed from now profitting even more from other peoples' work. But Adobe seems to care more about these pople than real photographers and illustrators anyways.

How much risk does a contributor actually face when uploading generative AI content?

I recently uploaded some AI images (nothing special, more to test if they would be accepted), but not before I had read the terms of the respective AI generators to see if I could use the content for commercial purposes. Furthermore, I adhered to all the guidelines that Adobe Stock imposes on gen-AI content. However, I did use generators that are currently being addressed by Getty Images and are not from one of the "ethical" generators (such as Firefly presents itself, aside from the fact that you can't use it for commercial purposes due to its beta status anyway). So, I tried to "play by the rules" as far as they are currently clear and thus attempted to limit my risks. But how do you view this? Aside from the question of whether you are in favor of or against AI content for stock purposes.

I do not think anyone can really answer the question at this point. There are a lot of lawsuits going on against AI generators that stole copyrighted content to train their AI, but as long as no single court judgement has been made (at least I am not aware of one being made yet), we are all just fishing in the dark. You say you try to "play by the rules", but at this point we all do not really know what these "rules" are. It's the very reason why most agencies refuse to even accept AI content at this point - Because the legal framework for it isn't really set yet.

Quote from: ribtoks on September 07, 2023, 04:48

It is actually strongly adviced to NOT use the exact same keywords for all of the images and this decreases their indexability. Yuri Arcurs was advising to use about half of the keywords the same and others - different. {snark removed}

advised by whom? is that a review for your product?  many images need 10-20 tags, and all highly relevant, so should we add different spam tags to make them different? and how do we know anything about how each agency does their indexing?

That was exactly my thought. If I have two images that are very similar and I want to tag them with accurate keywords how am I supposed to make 50/50 different keywords for both? It would mean that I have to leave away relevant keywords for one image and/or add irrelevant to the other image. I don't really see how this could work a different way.

Im still not seeing this.  I'm getting maybe a 2-3 day time period for new images to be uploaded and its been like that for a while.
Last batch i submitted on the weekend have been reviewed sometime before now when i checked.

That said, for some reason, ive had 4 images stuck in the queue for 2 weeks now that havent budged but the other 150 or so went through.

You are lucky. My oldest batch of photos (not AI, real photos) has been sitting in review for 19 days now, still not reviewed. The batch after that has been sitting unreviewed for 16 days, another for 10 days, another one for 9 days and so on...
The only photos that get reviewed faster for me are editorial ones.

But right now it kind of looks like my review queque has not moved at all in days.

Quote from: Her Ugliness on September 05, 2023, 04:26

All allowing AI images to be submitted as photos did was prolong the review time for real photos as well.
My real photos have been waiting to be reviewed for 16 days now. It really sucks when you want to submit time sensitive content.
I usually submit all my images to all agencies at the same time and managing different submitting times for individual agencies is really quite the hassle - especially since Adobes review time changes all the time  -  getting longer and longer.
Will my Halloween photos be reviewed in 2 weeks or 2 months? Impossible to know when the right time to submit time sensitive photos is now. 

You complain against flooding AI images, and, in the same time, submit Ai images yourself. You set fire to the forest and complain that it burns? Modern maturity... 

No, read again.
I am complaining about AI images being submitted as photos. They should never have been submitted as illustrations or photos in the first place. They are not illustrations. They are not photos. They should have gotten their own AI image category (or never have been allowed at all).
I complain about Mat claiming that Adobe does not need more reviewers, when very obviously they do and need to change something. They either need to disallow AI images, give a separate review queque to them, so people who still put the effort into real photos and real illustrations don't have to suffer even more disadvantages because of AI images than they already do or hire a hell lot of new reviewers.

For a second upload, include in the notes to editor "Previously submitted." 

Adobe does not have any option to include such a note to reviewers.
And I never understood what good the "previously submitted" option on Shutterstock does, other than telling a reviewer that another reviewer saw something wrong with the image and making him biased. If you ask me that's just making your chance of having the image approved smaller.

In a German article about what can artists do to protect themselves from ai, they mentioned �glaze�, a software that slightly changes your images and males them a lottle soft or fuzzy but renders them unusable for ai.

https://glaze.cs.uchicago.edu/

This is maybe something agencies could use to prevent the data scraping of our work without proper licensing of data sets.

And that we might use on our own websites.

Will try to learn more about this.

https://www.makeuseof.com/how-to-use-glaze-protect-art-from-ai/

Thanks, that's interesting. Sadly the article fails to explain why they think this will make it "near impossible for AI to copy your style". It looks like it just added slight artefacts to the images in the example in this article. So, why do they think this will in any way keep AI from training on your images? I don't understand it and the article fails to explain this to me.

And I don't think this is in any way suitable for microstock and photos. It's suitable for a personal artist's portfolio maybe, but when you actually want to sell an image, like microstock agencies do, you will want to sell a high quality version and not something with artifacts. Especially on photos this would be noticable. Any agency would reject a photo with such artefacts. And even if agencies used something like this for their previews - as soon as a customer buys your image and uses the version without the artefacts somewhere online it's "free grab" for the AI again.