|
MicrostockGroup Sponsors
This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.
Messages - Andrej.S.
Pages: 1 ... 4 5 6 7 8 [9] 10 11 12 13
201
« on: March 11, 2024, 03:54 »
Yes, it doesn't seem to be as easy as you might think at the first glance. I found one "study" with analyse of the keywords and other factors of the 10,000 bestsellers on SS: https://xpiksapp.com/blog/microstock-keywording-analysis/There are some interesting insights, some no-brainer but aswell some patterns you would not think about. Probably one should focus more on low-competition keywords by using more verbs and adjectives and less nouns. Aswell one should probably try more 2-word keywords. https://xpiksapp.com/blog/keywording-files-for-microstock/Another interesting aspect to think about is using various image descriptions and keywords instead of the same ones for series, since then images would not compete against each other. Aswell not using too generic keywords. I wonder if updating the keywords to current new trends would lead to a better visibility / rankings and more sells or if the ranking is already "fixed" within the time.
202
« on: March 10, 2024, 16:51 »
Interesting to read that many are currently similarly stuck at 80 to 90 USD.
Dreamstime states as of March 2024 there are more than 1.2 million contributing photographers.
Assuming there are probably 200 thousand contributors in the same situation like us, this would equal almost 20 million USD in unpaid revenues.
With the current high interest rates on the market this is a good sum to invest.
Sounds like a kind of business model.
Only if it is legal to invest the money. If it's illegal or they make bad investments, we're even more screwed. Investments aren't always profitable. I don't agree that it's a good business model and I don't think it's legal either.
Yes, legally and in accounting terms it would be most likely at least a grey area, as they are obligations. I think they have to account unpaid "outstandings of contributors" as reserves, comparable to pension liabilities for company pensions. On the other hand, pension liabilities for company pensions may be invested in safe investments such as funds, bonds, fixed-term deposits, etc. Currently you would get over 5% yield for 1Y US Bonds, what's not that bad if you think that S&P 500 is performing not much better. https://www.bloomberg.com/markets/rates-bonds/government-bonds/us
203
« on: March 10, 2024, 06:26 »
Interesting to read that many are currently similarly stuck at 80 to 90 USD.
Dreamstime states as of March 2024 there are more than 1.2 million contributing photographers.
Assuming there are probably 200 thousand contributors in the same situation like us, this would equal almost 20 million USD in unpaid revenues.
With the current high interest rates on the market this is a good sum to invest.
Sounds like a kind of business model.
204
« on: March 09, 2024, 03:36 »
Thank you for the information.
What does customers response mean?
How many customers how often viewed the image with the same searched keyword order of the image during the first 30 days?
It's currently a dumb thing.
They should replace the algorithm by letting the AI identify and describe the images and their content, so the customers could search more precisly.
/Edit:
Well probably the algorithm will remain a business secret.
Found various statements.
On the offical blog it is stated that titles are not used for rankings.
https://blog.adobe.com/en/publish/2019/06/18/stock-keywording-tips)
On the developer blog it is stated that they already use AI since several years to increase the variety of search results.
https://blog.developer.adobe.com/evaluating-addressing-position-bias-in-adobe-stock-search-9807b11ee268
You are welcome and thanks for the additional.
"Do not spam keywords irrelevant keywords can actually hurt your visibility on the Stock site."
The whole article about position bias was interesting. Yes, Titles are not searchable in Adobe Stock.
I assume you mean the same thing when you say image rank as search discovery? But not searchable means much more. If you have a word in the title and not in the keywords, it's not going to be seen in a search for that word. Which images come up on the first pages, is something different.
"What does customers response mean?" yet another mystery in the secret system that determines image rank. We don't know, they aren't telling. No agency gives up their "secret sauce" recipe. The assumption that it's views and sales, or how many related word matches, is like saying, to make soup, you need to add water.
Everything after the basics is what makes the search what it is. Those basic starters aren't the intricate and complex answer. 12-13 years ago, you could pretty much see that IS and SS were based on the time on the system divided by sales. On IS a new image with one day and one sale, would leap to the front. Then after a week, no more sales, it would start to drop and after a month, if no more downloads, would fall back in the general pages.
SS until they fixed the search (a view depending on who you are) to avoid some of the bias, someone who had an image up since the first years of SS, with lots of downloads, would be on page one, and the rest of us,could get into that page. Kind of a king of the hill situation, where top images, stayed on top, and got more downloads, because they were on top.
Sometime years back, they changed everything, you might find the threads about how "I've been on page one for years and now I'm on page 12!" That's life. The favorites with an unfair advantage, got adjusted.
The search changes, or can, by location, season, specific keywords, simple one would be, St. Patrick's Day will be higher now and Easter will get a priority. In a month, they will drop down and some other events or holidays will be raised and favored. Everything is always changing, so how would someone know what it is? 
If you go to church every day and pray that your images move high up of the algorithm they will eventually rise. If you pray, you will be listened. That's what I was told since I was a child.
Yes, I admit, I think that's intended as humor and it's also true. Since we have no control, we can pray, or burn incense, or sacrifice chocolate chip cookies to our inner sales hunger. The answer is, best keywords, best images, upload quality with proper tags, sit back and hope for the best. Worrying, conspiracies, trying to play the system, and many other side efforts, are mostly useless and a waste of time and energy.
You are absolutely right, I didn't even take seosonal search patterns into account. These will certainly also influence the image rankings. I think it's a shame that Adobe doesn't show the daily top 100 or at least top 20 search pattern like google does it in the search bar. Or top searches of the week, month, etc. Would be very useful and you would save a lot of time instead of trying out which content is in demand or only guessing if you have the right keywords.
205
« on: March 08, 2024, 17:41 »
One thing I find funny is people need to use words correctly. The "training" is "stealing". Call it for what it is. The "ai" (which is not true ai, but essentially sophisticated theft & pattern re-arrangment) STEALS images. They need data to STEAL. If it doesn't have data to steal, then it can't "train".
Is training stealing when they pay for the image use?
Obviously not, but you can argue about the fact that some agencies like Adobe or Shutterstock either don't let or delayed the opt-out for contributors. But yeah that's business. Read just on Adobe's Discord channel, that the staff of Stability AI have tried to harvest (steal) all prompts and images on Midjourney, so that their servers collapsed for 24h and Midjourney banned all Stability's staff accounts. But can't tell if that's really true. Would be crazy.
206
« on: March 08, 2024, 10:33 »
(a) It's REALLY important to remember "ai" tools are based off of THEFT of assets. The biggest "problem" "AI" companies have is figuring out how to remove WATERMARKS. I.e., because they STEAL assets.
(b) It's important to push for/request with companies you deal with that you be compensated for the stolen assets. It's very easy to 'backwards compensate' people - it requires some programming - but definitely doable.
(c) Authors should be compensated in perpetuity, because the "ai" tools (goal) is to sell in perpetuity.
Why do you think that artificial intelligence steals things?
, they wanted to pay for it, they wanted to make a semblance of compensation!!!
They wanted to clear their conscience and avoid legal proceedings, but in fact, it is a mistake that they made, because by doing so they recognize their great accountability.
The developers say that this intelligence is like the human brain. He is taught by the example of some works, but then he creates his own works. You also learned to draw or photograph from the example of other peoples works and copied them at the beginning. Then you started creating your own works. So is artificial intelligence.
Have you ever developed a skill in your life? What is your competence if we unplug the machine? The AI did not scrutinize reality to learn, it ingested the work and expertise of real people. His vision is not objective, it is oriented and inspired by human work (which has been copied/looted). And human vision is driven by emotion, real organic intelligence, not a machine vision.
Come on. This is just ridiculous as an argument. What's your skill, if the supermarkets will be closed. Can you grow potatoes in the field, milk cows and make cheese?
Or can you sew your own clothes?
These are all skills that only a few special ones do today because of the industrial revolution. The same will happen to photographers. Why do you think that a skill of a photographer is much more worth than from a farmer? Because it's based on emotions?
Knowing how to do something does not mean knowing how to do everything, your arguments are really ridiculous.
As of tomorrow, I am fully capable of managing my food autonomy; I have developed this skill over the years. And I have quantities of organic, non-hybrid seeds...
You can well imagine that machine learning was done from the work of competent people, having produced quality images that sold, the machines did not ingest the unsaleable crap produced by those who today feast on generation by AI...
Ok, probably this conversation will get ridiculous, but I'm gonna give a try. Do you mean that you or the humanity is capable of managing your food, for example by planting seeds? If you are that skilled that you can grow own food and have a garden, then hats off to you! I speak 2 native languages (german and russian) and speak somehow ok englisch and a little bit french (as kid in school it was much better), but really I would probably starve to death, if there would be a crisis on earth and the food supplain chain would not work anymore for several years. Or probably I would simply raid the farmers' fields so that my small daughter and my wife can survive. Human expertise and competence are indeed extremely necessary but you have to decide which profession we are talking about. Is it a highly qualified profession like a hand surgeon or a farmer, or a 9 to 5 office armchair man or a micro stock photographer. I mean, yes, I'm sure there will still be extremely highly skilled photographers in fields like wedding or architecture who can still compete against AI. But there won't be any masses of them.
207
« on: March 08, 2024, 09:35 »
(a) It's REALLY important to remember "ai" tools are based off of THEFT of assets. The biggest "problem" "AI" companies have is figuring out how to remove WATERMARKS. I.e., because they STEAL assets.
(b) It's important to push for/request with companies you deal with that you be compensated for the stolen assets. It's very easy to 'backwards compensate' people - it requires some programming - but definitely doable.
(c) Authors should be compensated in perpetuity, because the "ai" tools (goal) is to sell in perpetuity.
Why do you think that artificial intelligence steals things?
, they wanted to pay for it, they wanted to make a semblance of compensation!!!
They wanted to clear their conscience and avoid legal proceedings, but in fact, it is a mistake that they made, because by doing so they recognize their great accountability.
The developers say that this intelligence is like the human brain. He is taught by the example of some works, but then he creates his own works. You also learned to draw or photograph from the example of other peoples works and copied them at the beginning. Then you started creating your own works. So is artificial intelligence.
Have you ever developed a skill in your life? What is your competence if we unplug the machine? The AI did not scrutinize reality to learn, it ingested the work and expertise of real people. His vision is not objective, it is oriented and inspired by human work (which has been copied/looted). And human vision is driven by emotion, real organic intelligence, not a machine vision.
Come on. This is just ridiculous as an argument. What's your skill, if the supermarkets will be closed. Can you grow potatoes in the field, milk cows and make cheese? Or can you sew your own clothes? These are all skills that only a few special ones do today because of the industrial revolution. The same will happen to photographers. Why do you think that a skill of a photographer is much more worth than from a farmer? Because it's based on emotions?
208
« on: March 08, 2024, 09:25 »
(a) It's REALLY important to remember "ai" tools are based off of THEFT of assets. The biggest "problem" "AI" companies have is figuring out how to remove WATERMARKS. I.e., because they STEAL assets.
(b) It's important to push for/request with companies you deal with that you be compensated for the stolen assets. It's very easy to 'backwards compensate' people - it requires some programming - but definitely doable.
(c) Authors should be compensated in perpetuity, because the "ai" tools (goal) is to sell in perpetuity.
Well, yes you're right, if you are considering AI image models like stable diffusion and Midjourney. The developers have used many copyrighted material and even had AI tools, which remove watermarks. But agencies like Adobe, Getty and Shutterstock understood the copyright infringement problem and are now creating either own AI Image generators and they currently "compensate" their contributors with a few crumbs or are selling or licensing their contents to the big AI companies. So it is just a matter of time when they will have tools that are on the same level like the current leaders like Midjourney and Stable Diffusion and they will be fine with the laws or even OpenAI will take the lead with their further developments, since they got the license from Shutterstock. I believe the copyright problem will be a short obstacle.
209
« on: March 08, 2024, 06:53 »
Dear human Colleagues,
As AI technology continues to advance, its understandable for us stock photographers and videographers to feel somewhat concerned about its potential disruptions to our revenue streams. We can just keep doing the same old and hope that it just blows over (which it wont) or we may adopt strategies to protect our livelihoods.
I've published five such strategies that you may wish to start adopting right away lets get started!
https://brutallyhonestmicrostock.com/2024/03/07/future-proofing-creativity-5-strategies-for-stock-photographers-videographers-in-the-age-of-ai/
Alex
Thank you for your sights and ideas! Interesting article! All strategies make sense, although I would say most portraits are already photo-realistic, so I would say one have to focus on new creative concepts to stand out from the masses in the future and can't be copied easily because the mass don't having the skills yet or the AI not enough training materials to recreate. Here one example for the portrait with the elderly man. Prompt with Stable Diffusion XL + AI Upscaler: An elderly man in a flat cap and suit looks into the distance. He is standing in front of a lush green field with rolling hills in the background. https://ibb.co/N23HqZkSecond try with same set up and following prompt: An elderly gentleman wearing a flat cap, overcoat, and tie stands in front of a lush green field. He looks off to the side, with a small hill in the background. https://ibb.co/5knKyjzThere are some generation artifacts and errors but you have to look closely and can retouche them easily. Hands are really tricky currently. But it is just a matter of time when you will create shots of hands without errors. You can already create some simple compositions with hands, although you still can obviously spot huge anatomy errors. Prompt with Stable Diffusion XL + AI Upscaler: A close-up of two beautiful hands, one male larger and one female smaller, interlocking fingers against sandy beach. https://ibb.co/gJF431BBut probably there are already fine tuned models, which I don't know yet. I would not focus on these type of shots. Complex nature landscapes are really hard to generate. Can't imagine there is already a good model. This content should be safe like architecture for several years. Although I have recently found an extremely good model for modern, stylic interior shots. This model generates only a few errors. So who knows what you can already find, if you search for a longer time.
210
« on: March 08, 2024, 05:07 »
https://brutallyhonestmicrostock.com/2024/03/07/future-proofing-creativity-5-strategies-for-stock-photographers-videographers-in-the-age-of-ai/
Input: An array of Mexican dishes is beautifully presented for a feast. The table is dressed with a colorful tablecloth, making the traditional cuisine such as guacamole, salsa, and cheese-topped entree even more enticing. Various sauces and fresh ingredients complement the central dish, suggesting a celebration or a special meal. feast,guacamole,salsa,cheese,traditional dishes,colorful tablecloth,food,meal,celebration,banquet,cuisine,tortilla,avocado,lime,cilantro,beans,rice,table setting,dinner,lunch,culinary,culture,mexican,festive,ingredients,entree,bright,vibrant,tasty,assortment, top down
Not every buyer will be able to write such texts for programs that create photos or videos using artificial intelligence. The buyer must also be a writer and screenwriter!
Nah, in future you will not have to write such long prompts. It's really naive to believe it will stay the same. It will become much easier with multimodal AI models. You would just chat with a LLM bot and promt: "Hi ChatBot, how are you going? Hope fine, so listen please. I need for the next food advertising campaign a photo of an array of Mexican dishes which are presented beautifully and mouthwatering for a feast. Can you give me please 5 variations? Thank you!" The AI will do the rest work. I mean come on. The current AI models are developed to solve complex professional demands in business industries like chemistry, health care, automotive, engineering, etc. Do you think they will fail with much easier generic content?
211
« on: March 08, 2024, 03:34 »
https://petapixel.com/2024/03/07/recreating-iconic-photos-with-ai-image-generators/
None of the copies are great, but they're clearly more than "inspired by" the very famous images they were trained on.
Actually they are really bad copies. I don't understand why you would want to recreate old famous photos. It's like recreating Mona Lisa, why should you? What's the point? Famous unique vintage content holds it's value in it's authenticity. But you can create new original images with new perspectives, new techniques, styles, etc. You would try to evolve and not stagnate.
213
« on: March 07, 2024, 12:01 »
Today I played a little with a combination of Stable Diffusion XL and an AI upscaler. I mean, considering that we are still at the beginning of the development cycle and I am just a newbie, who plays with AI tools for 3 months the possibilities are already now promising for the coming future in the next years. I believe that future AI image models could be structured as "multi-layered" models. In the first step a rough image sketch with a few diffusion steps (almost like a composition sketch) would be generated based on the prompt and subsequently each identified content of the image would be filled in with more details by using fine tuned AI upscalers. So you will for sure have fine tuned portrait upscale models for skin, hair, lips, eyes, hands, fingers, etc. And probably also for other type of motifs, which will work like "inpaints". Like a painter, who would add details to an oil painting by several following steps. This direction is already being explored with models like Stable Cascade. I think the current diffusion technique is still very ineffective because of the current hardware limitations. Certainly there will be many new techniques and model refinements in the future which we can't even imagine at the moment. I also believe that the current generation errors will be spotted and corrected by the AI automatically. So in a nutshell for me, there is no longer any uncertainty that (stock) photography will be replaced by AI. The question is not if, but when it will happen. Some following examples. Generated with SDXL: https://ibb.co/9VwFMXSAI enhanced: https://ibb.co/pWZs2G9Generated with SDXL: https://ibb.co/H4skKNCAI enhanced: https://ibb.co/b5gZ5QvGenerated with SDXL: https://ibb.co/4FtjrL0AI enhanced: https://ibb.co/RNjqyDfGenerated with SDXL: https://ibb.co/KxrRPmSAI enhanced: https://ibb.co/GHNT1hQGenerated with SDXL: https://ibb.co/HPC9Lm3AI enhanced: https://ibb.co/0XtDm3zGenerated with SDXL: https://ibb.co/HHdjDJKAI enhanced: https://ibb.co/9mxFNt9
214
« on: March 07, 2024, 07:19 »
Adobe Stock clearly state they give a boost in sales to its active contributors, meaning that if you upload 10 images a day for 20 days, you will get a boost over a contributor who uploads 200 images in one go and then stops uploading for the next 19 days.
Really, they say that? Where? What makes active? One a week, ten a week, ten a month...
Just read at the Discord channel. They really encourage users to make small batches but without real reliable arguments. They say it would make more sense since you would then test, if the content pass through the review and is in demand and then upload more. But probably they just want avoid spammers.
215
« on: March 07, 2024, 07:14 »
Same here. Currently waiting at 85.78 USD. After the payout I will delete all content and delete the account.
20 or 25 USD would be acceptable. I would even accept a small payout fee like paypal does since you would be save not loosing bigger amounts if they become insolvent.
216
« on: March 07, 2024, 03:22 »
Just wondering if people with a higher acceptance ratio find their images higher up the algorithm?
Just a thought
Its not a higher acceptance rate, but your weekly standings score that pushes your images to clients.
Weekly Standings mean nothing. They are only for our own information.
Where did you get this information, that they are artist or image rank?
Because after a successful week all of the sudden old images that I never sold before are somehow visible in searches and get sales. What changes? Definitely not my acceptance rate. If not successful ranking, what do you think it is then?
Do you know how Adobe decides whose images to show in searches?
One of my new Easter bunnies image with 1 sale is on page 3 now. So search is not only based on sales of that particular image or keywords/title -
I don't claim to know how ranking works for the search, sales, or matches, or featured because of the season, but I do know it's NOT based on your artist rank. Did you see what Mat posted?
No.
-Mat Hayward
Did you watch the webinar about image rank and how it works? There were two of them.
After both webinars, here's what I think they have told us. This is after listening to the discussions of all three people from Adobe and making text notes.
In Order...
1) Images initially get ranked by customer response during the first 30 days. Your keywords and title are most important during that time.
2) After 30 days, changing the keywords or order, will not have much effect on image rank, from the customers. There are other factors that still can move an image up or down.
3) Adding detailed information, location, keywords or concept word combinations, will still help get an image found. Updating older images may not change the rank, but it will make the image more searchable.
4) Categories are not very important
Repeated many times, and the way I have viewed keywords and titles. Please understand these are paraphrased quotes, not exact word for word. But they represent what Judy and Mat said:
Only use relevant keywords you don't need to have 49 keywords, only the appropriate keywords. "If you were looking for this image, would you use this word to find it?" "If you searched a word, would you expect to see this image?"
"Is this something I'd expect to see for this search term."
If customers are looking at images and they are not a good match for the words included, the rank will be dropped. Anyone who includes words that are not relevant, will be hurting the image rank, and that rank will be attached to that image, pretty much fixed at that rank, after 30 days.
Watch Please? This comes from Adobe not someone who just thinks they see something. Adobe Stock search
Hi Everyone,
I've scheduled a webinar with our Senior Product Manager in charge of search this Thursday, February 27 at 2:30PM PST. It would be great if you can attend. We'll do our best to take as many questions as possible during our 45-60 minute conversation about Adobe Stock search. How you can use the tools to your advantage as a contributor at Adobe Stock. I hope to see you there!
https://www.crowdcast.io/e/search-tips-and-tricks
Mat Hayward
Thank you for the information. What does customers response mean? How many customers how often viewed the image with the same searched keyword order of the image during the first 30 days? It's currently a dumb thing. They should replace the algorithm by letting the AI identify and describe the images and their content, so the customers could search more precisly. /Edit: Well probably the algorithm will remain a business secret. Found various statements. On the offical blog it is stated that titles are not used for rankings. https://blog.adobe.com/en/publish/2019/06/18/stock-keywording-tips) On the developer blog it is stated that they already use AI since several years to increase the variety of search results. https://blog.developer.adobe.com/evaluating-addressing-position-bias-in-adobe-stock-search-9807b11ee268
217
« on: March 06, 2024, 12:06 »
The big question will be if big agencies like Adobe really want to sell generic content from the contributors in the long term. They would have still to pass on contributor's commissions of approx. 30 to 40%.
I doubt it. Look at Adobe's Max sneaks. They have developed so much AI stuff and already are integrating it in their tools.
They have developed even an own LLM.
The invested for sure a huge amount of money. They have to rearn this amount quickly to get joyful investors.
Firefly is getting better and better.
They stated they had already 3 billion generations with firefly in couple months.
And look what they are trying with offering "Mission" contests in the community to finetune their image generator for just 50 to 80 bucks.
When their models can replace most of the current content, they will just recreate the majority and take contributor's content down from their servers.
They will find people, who will contribute some stuff they need to finetune their models for some bucks.
Or they will only allow inpaint modifications in firefly for only with their tools generated images and movies.
This will happen in the first step with photos and then movies.
Just watch the developments in the following years.
218
« on: March 06, 2024, 10:25 »
Depends on how wide/narrow is the acceptance of the word "subject" here. The real question is: how similar are the images? If you have 200 images with, say, the theme of healthcare, or financial sector, you should be ok with submitting them all at once, provided that they are not similar. (I can easily think of 200 different images depicting each of these topics.) On the other hand, if you have 200 images of a very specific subject, chances are they are too similar, and you are better of not sending all of them, not even in multiple batches.
Assuming you have images that are different enough, I would still send them in multiple batches spread over multiple days. Adobe Stock clearly state they give a boost in sales to its active contributors, meaning that if you upload 10 images a day for 20 days, you will get a boost over a contributor who uploads 200 images in one go and then stops uploading for the next 19 days.
The boost reward doesn't make any sense. Never have seen a difference in my sales. There are so many professional high end photo sets, which contain many hundreds of similar photos. For example electro mobility with shots showing people charging cars or real live situations with family members. You wouldn't upload them by splitting in 10 images for a single batch. You just lose productivity because Adobe's upload system is extremely user friendly for huge batches of similar images. Since review goes up to over a month you still accumulate images pending in review. Just upload them at once.
219
« on: March 05, 2024, 11:04 »
Just found the research section on Adobe. https://research.adobe.com/publications/Have you seen guys what Adobe has been working on lately with scientists like MIT? As I said some time ago Adobe should focus more on AI, lol they've probably been doing that already for years. Docens papers on AI research, for example extreme speedup of diffusion models (reduction of the generation time to 80 milliseconds with just one generation step): https://research.adobe.com/publication/dmd/https://tianweiy.github.io/dmd/Approximate Caching for Efficiently Serving Diffusion Models, which reduces aswell generation time by using caching method. https://research.adobe.com/publication/approximate-caching-for-efficiently-serving-diffusion-models/Or Iterative Multi-Granular Image Editing Using Diffusion Models, which means you can edit image content by just text prompts. https://research.adobe.com/publication/iterative-multi-granular-image-editing-using-diffusion-models/https://openaccess.thecvf.com/content/WACV2024/papers/Joseph_Iterative_Multi-Granular_Image_Editing_Using_Diffusion_Models_WACV_2024_paper.pdfCoPL: Contextual Prompt Learning for Vision-Language Understanding, which means developing a user friendly and precise prompting technique. https://research.adobe.com/publication/copl-contextual-prompt-learning-for-vision-language-understanding/Perceptual Artifacts Localization for Image Synthesis Tasks, which mean an AI model to identify artifacts in AI generated images. https://research.adobe.com/publication/perceptual-artifacts-localization-for-image-synthesis-tasks/You guys are counting on Adobe for nothing. It was such a brilliant idea from Adobe to accept AI images. They have millions of user generated images and there is still a flood of new coming millions, which they can use for their own research purpose. Just imagine, if they combine the AI technique of spotting generated artifacts, which they can correct on the fly. Dam*, their strategy is just brilliant.
220
« on: March 05, 2024, 05:18 »
I would say 5 to max 7 years is a realistic scenario where only the big agencies like Adobe Stock, Getty / Istock, Alamy, etc. and of course agencies with editorial content will survive.
Perhaps small very specialized agencies (food, architecture, etc.) will survive by license supplying AI model developers.
All others are either too unknown, poorly diversified or have no recognizable long-term strategic orientation. They will disappear.
3 years are to short to see large market changes because of current AI technological barriers. With the exponential technical development, the marketing and thus the awareness of AI models will increase significantly the following coming years. In addition, many users most likely have medium-term plans for stock licensing, so they won't switch next year. I think we will see a strong shift for the first time in 3 years, when image generation will be possible in real time.
In my opinion, there is a scenario in which the big stock agencies could remain profitable in the long term without ending up like Shitterstock as a pure data supplier for AI model developers.
You would need to develop your own AI assistant that allows prompting that combines an AI model with stock or customer's own photos using img2img with inpainting / outpainting.
This would have the advantage of being able to fulfill a wide broad of customer wishes through a high degree of flexibility by using the available high res photos with very low image errors.
You have to get away from the rigid old concept of a pure image database.
I have a design suite platform like Canvas in my mind but with focus on images, 3d and videos.
Let's take the example of a small car repair shop that wants to create a customer offer for cheap winter tire and oil changes as an advertisement for Christmas.
The marketing manager has an in oil covered smiling Santa in mind who is changing the tires in the car shop garage.
He could either take his own photo of his own garage and then pick out a Santa and integrate it into the photo in real time using an inpaint command, or he could use a prompt command to display a garage and a Santa separately, pick out suitable ones and then merge them into one picture. He could leave copy space free for text or his own logo, etc.
The advantage would be the extremely high individual flexibility in comparison to current AI image generators.
Agencies need to integrate AI in a clever way.
There is no way around it. They have to invest money to remain profitable in the long term. And only the big ones can do that.
I would bet heavily on Adobe Stock to expand in this direction.
They actually have a lot of expertise in this area with their software design suite.
So integrating stock photos into Photoshop was a smart first step. But Photoshop is too complex for the normal user so they have to replace it with an AI assistant.
People will need no agencies for that, it will be free applications available that will do everything for free. SS is charging a customer for something the customer can get for free in many applications today. Imagine in one or two years.
SS is a sinking boat like most of today available agencies.
Well, I don't think such AI assistance tools will be available completely free, that's not a viable business model in the long term. There would be also no API connection to the databases. I think agencies could market their inventory better with AI. About 80% of their inventory is completely unused and hasn't been bought once. This is often because keyword searches are extremely imprecise. I mean there are already hundreds of millions of images that are just rotting away somewhere on the server because often they just don't show up at the top of the search rankings, even though many of them could also be in demand. They should develop like Adobe more into a cloud-based solution with an API connection. But actually I think the big ones will just promote their own generators because they don't have to pass on any commissions to the contributors, so they can earn the maximum. Probably they know themselves that most buyers don't need outstanding image quality and huge image resolutions. Like IStock. Just look how they even created an user friendly promt builder with various aspect ratios and inpaint possibility: https://www.istockphoto.com/ai/generation/aboutIt's like I already said many times before, the prompting will become so much easier in the future.
221
« on: March 05, 2024, 03:24 »
Anyone else's review times are screwed up now? Today, in my to-be-reviewed tab, my oldest illustration now say "submitted over a year ago" and my oldest photo "submitted over 2 months ago".
Same here. Many of my images, which are stuck in the review are now shown submitted over a month or two and some even over a year ago. I hope it's not due to the AI image flood, although it would be obvious that they get review problems every now and then.
222
« on: March 05, 2024, 02:39 »
This AI software generates content in the cloud
This means you will pay a lot of money to the cloud and pay constantly for every video created.
It will be cheaper for a buyer to buy a video on stock than to pay to the cloud.
Where are you getting this information from? Please provide comparative costs for purchasing a stock video versus generating an AI video.
C'mon he is either delusional or just a web troll. You can't expect to get valid arguments. Haven't found any details for the generation costs but there are rumors for a subscription plan of 60 minutes generation time for 20 bucks and 500 minutes for 50 bucks. Aswell a pay to go model with costs about 0.01 to 0.10 USD per second. But I am quite sure the prices will drop fast since NVidia is investing billions in new AI chips for real professional demands like in health, chemistry, etc. industries. So you could rent in few years very advanced GPUs, which should shorten heavily the generation time.
223
« on: March 04, 2024, 14:55 »
Hmm, I try to explain my idea more clear.
Yes, it would contain "mass spamming".
If the agencies would accept a lower image resolution for AI images, like for example 1 mpx (what you would call approx. small to medium size res.), which is enough for mobile content, then they could decrease the image price to 15 cents for buyers because it would take just about 3 to max. 5 minutes to generate and upload an AI image whitout need to upscale and retouche artefacts.
Hell yeah, I would then spam the s*** out and probably even automate the image generation and upload dozen thousands of images a month. I don't understand the whole effort. We currently upscale and retouche to pass the review although these buyers need just the smallest resolution. It's like casting pearls before swine. They just need gap fillers for their websites, news articles or blog / social posts. These buyers do not appreciate anyway the quality of good photos.
One should even consider saving the review staff team for this content.
On the other hand I would try to do a lot of marketing on classical high quality content especialy for max image resolutions, which is needed for print demands. I would increase the prices to min. 5 USD but better just remove them completely from the subscription model.
I mean, let's take Adobe Stock as an example. What is their long-term strategy right now? Simply let themselves be flooded with AI images and watch as the market loses value inflationarily?
Their premium collection is practically nonexistent and too unorganized. You have to separate buyer segments now to position yourself against the flood of AI images. Because in the long run, current AI image buyers will most likely transition to AI image generators as part of multimodal models in a few years anyway. So they would have just devalued their remaining authentic content as well.
@cascoly
Yep, it is a also review problem. AI can't currently compete if you consider native high resolutions. Can't imagine that it will get so much better in 3 to 5 years.
I would have probably sorted out at least 90% of the AI content because I know that upscaling inevitably creates artifacts that need to be painstakingly retouched.
So it just does not make any sense to want the same image resolution for this content.
224
« on: March 04, 2024, 09:09 »
Hell no. It is a lot easier and much faster to take a good series of pictures and videos with a camera than to do stuff with ai.
At least for me.
Sure you can do many very simple shots in much less time. Of course, I can photograph myself brushing my teeth in the next 3 seconds. But I would say you are very limited to specific motifs. For example to take photos of easter eggs one is already dependent on weather and time. You need a somehow a beautiful well-kept lawn, nicely painted Easter eggs, etc. It's a motif I would always generate because I know there would be only few artifacts and I don't need an authentic shot.
225
« on: March 04, 2024, 08:55 »
From the cotributors point of view it makes sense, but from the customer's point of view it doesn't make sense. For them only the end result matters. They want an image that suits there need and it doens't matter to them how much work the image was to the contributor.
Even now different photos take so much more work. I have photos in my port that took me literally 5 seconds to shoot, and photos where I spent lots of money on props and easily 20 minutes setting up the scene, one hour taking the photo (especially with pets that you can't just place easily like you want and often have to try over and over again till you get what you want) and 20 minutes post-processing. Yet both images are sold for the same price. It doesn't really matter to the customer. If you want to be paid per time/effort, then microstock, where everything is offered at the same price, isn't the best business model for that to begin with. It's not like assignment jobs, where a customer tells you what he/she wants and you say "This will take me x hours and therefore I will charge you X amount of money".
Yes, the props and an aesthetic composition, lighting, etc. cost an enormous amount of time in food photography, for example. With animals, as you've already written, you need a lot of shots, which you have to sort out after the photo session. So it takes for the majority of shots really a lot of time. First users naturally focus on the price. Many for sure expect everything for almost free, especially for mobile use like in newspaper articles or blog posts, etc. On the other hand I am sure there are designers who need high-resolution content for print projects as well and would pay more, if they would get a set that maintains continuity (several shots) of the same motif. But currently, everything is being mixed together. Agencies like Adobe should differentiate much more now; they are currently contributing to an inflationary price decline due to the high flood of AI images. In the long term they can't anyway compete even with lower prices against AI image generators, which the first users will use for mobile usage. One possibility could also be to accept AI-generated images up to a smaller certain resolution, for example, 2 megapixels, and offer them at a particularly low price for mobile usage only. Since the AI images almost always generate artifacts when upscaled for higher resolutions, this would help alleviate the review process. I believe it's primarily a marketing problem and higher prices could be achieved by better hand-picked authentic, genuine photos and marketing them in various premium galleries. I mean look at other smaller agencies, they can't be real competitors because they aren't performing good.
Pages: 1 ... 4 5 6 7 8 [9] 10 11 12 13
|
Sponsors
Microstock Poll Results
Sponsors
|
Show Posts
