Nice, that was an interesting breakdown ! I knew about the loudness war but not about all the finer points brought up about studio recordings. Funny how it all comes down to things being fresh and original when made within constraints as opposed to using infinite computing power to just do the same thing over and over again.
I've personally been enjoying Chromeo a whole lot for years now. Of course they mix everything digitally and are highly technical (they even have a very interesting video dissecting one of their tracks) but a certain analog feel still seeps through, and apparently they manage to perform great live sets for a 2 men band.
First let me say this is a fantastic thread with a lot of in depth breakdowns and thoughts from many areas and for that I'm very grateful this discussion exists here.
There are many problems the with the entertainment industry as it stands now and I feel it's difficult and fuzzy to point to any one area and shout THERE"S THE PROBLEM! I'll attempt to share a few thoughts that are loosely connected with the hope of forming more coherent thoughts on the matter.
The "sameness" was a trend long before we got to AI, in fact I think it may be an inescapable principle of scaling in general. I saw a research study in the early 2010s that showed that if you take portraits of several groups of people from around the world and split them into different sets of images, then average the sets of images together the result is that each set produces roughly the same face no matter the widely ranging individual looks. I felt this hinted at a larger principle as it described what I was starting to see emerging in the games industry. Fast forward to 2024, While we have a very wide array of indie games, anything AAA has trended towards Pixar if it's stylized or towards hyper realism if it's something like COD. I think AI accelerated this 10x as that seems to be what it's good at. (accelerating what's already there) I really do feel like overall most things in production or recently reduced can be described as "soulless". My kids watched the Mario movie and all I could think was, this is fun but it has no soul. We're in a time of transition and the role of the artist won't end with us, but it will change radically. Many of the things going on today need to change, and I think we all feel it. I don't think all the lawsuits will stop the progress. If we look at the automotive factory workers from Detroit I think we can all see where this goes. We still have automotive factory workers, even at Tesla. But the job has changed drastically. I'm not sure I will like the new world but I'm genuinely not sure what to do about it.
I do think we will see small shops pop up with all human crafted material at a premium but that will be a very competitive market with only a few winners.
I don't get that comic. Does it predate the big explosion in AI image generators? Because if not, it doesn't really make sense when the AI it features has no relation to the actual ones.
The tool will be exclusive to Pocoyo and any licenses Animaj acquires or partners with, though they will release some tools to the public.
Sounds like a scam. These AI aren't content aware at all, let alone on the level you'd need to be to understand a storyboard and transfer that into actual 3D movement, especially filling in all the stages inbetween. Doesn't convince me when, without signing in, 90% of the comments are people who call themselves "angel investor" or "generative AI developer" or such. And I mean...If I wanted animation, I'd go for an animator.
There are many problems the with the entertainment industry as it stands
now and I feel it's difficult and fuzzy to point to any one area and
shout THERE"S THE PROBLEM!
It's simple. Financial clout has crapped all over the copyrights of creatives. It's that basic. No amount of legal spin smoke screens can justify what has been allowed to happen. Its flat out reprehensible theft on an industrial scale.
Thanks for posting that. I have to say this is a great use of AI. There's still a bit of AI jank happening on that face. Though it's much less uncanny than the early days of CGI characters and as they say "This is the worst it will ever be." I get the feeling that this AI tooling is moving in the direction of being locked away in abstraction for the vast majority of game developers with Nvidia and other big companies offering the masses a turn it on or off option when using it to develop a game. Maybe that's good, maybe that's bad and maybe I'm wrong. Just thinking out loud here.
Like shown in the "hyper realistic..." video, the ai character looks just different and less detailed, just look at the eyes and make-up. The other shown technologies are much more interesting imho.
Good or bad? "art engineer", what "he" will create next... 4k followers, person doing art for what idk 20 years 10 or less followers. winning, stealing seems to be winning here="your earth"..
wasn't looking for it, not promoting them at all quite the opposite, just showing that people don't seem to give a sh!t either way till it affects them.
The only "decent" usefulness i see for... this idea, for all aspects of employment is, C&C = when you want quick c&c and YOU make the adjustments not a script.. i will still not like it no matter what, especially when people make money off it, that is just a punch in the face to any creative with any employment.
They tend to be very tech-forward, but I'm not sold on AI actually replacing whole persons. It can work great as a tool, but it needs constant guidance from a human to correct errors.
re-ai code generation.... I was playing with around with Deep-Seek the other day.... It comes across as a "Bullshitter".... it talks a good game, all the right buzz words and superficially the code it generates looks kind of right but it's only skin deep, if it would compile it wouldn't do what it was supposed to do.
Also, lets say any of this did work and we went 5 to 15 years into the future for the "next-gen" to start "producing", if something happen oh idk like they (the ai bros) decided to pull these things or ask for stupid prices.
The next gen would all be screwed as in, where do they start if they wanted to produce?
Not having the foundational basic knowledge is going to stunt everyone in everything, that is a pretty big deal i think, now imagine if it took over all aspects of our "societies", giant red flag, no?
How do i do this thing, then those books for dummies comes to mind rather quickly.
But yeah the buzz word is the thing, i guess this is some sort of scheme by everyone doing these things to scam big money makers $ with too much money into "plugging", into these things and then quietly pulling the cord and saying oh well its just a loss for you it didn't work out, kind of vibe i am getting.
Gotta love that the input really doesn't matter much for the output? I mean that end result is nothing like the input, besides yeah some toony fox and racoon characters. Can anyone even claim to be in control here, art directing with such a gamble is... Kinda pointless Unless it all just does not matter
Gotta love that the input really doesn't matter much for the output? I mean that end result is nothing like the input, besides yeah some toony fox and racoon characters. Can anyone even claim to be in control here, art directing with such a gamble is... Kinda pointless Unless it all just does not matter
Thinking thats the gist with all generative AI, most AI promoters really look at its development and output as something to hustle with. This one seemed a bit different in its approach, though probably comes down to how people use it.
The usual gacha AI approach has a market, just not sure its something that can be sustainable when it comes to monetization.
I have been trying to use AI tools in my workflow and it's just not worth the effort, at the moment the tools just don't work, I can make the thing faster myself and knowing it will not be a mess that I have to clean up after. I keep reminding myself that there are a LOT of BIG players that have a lot to loose if AI suddenly is capable of doing anything a professional Artist can. These are generalizations , I understand things are way more nuanced than this:
No need for 3D Modeling = no need for Maya, Max, Blender, C4D and so on. No need for rendering / texturing = no need for Vray, Redshift, Substance etc
Of course these companies are doing their AI work to incorporate it into their software and "stay relevant". But I honestly just don't see it happening. At least not in the coming 10 years? AI slop is that, slop, it's boring, it always has "that look", and it's just tacky.
I sometimes compare it to the "not everyone should have a podcast" thing. Just because you can does not mean you have to or that you will create quality things.
The democratization of tools is there, and that's awesome, but just because you can now "make an animation with a prompt" does not mean that a) it's going to be good b) people will actually consume it let alone PAY for it.
I don't know, I'm just confused at this moment being a 3D Artist, I'm keeping myself busy, working and learning but holy crap the AI ghost creeps in frequently and I hate the feeling.
never understood how AI is considered a 'democratization of tools' tbh. Smells like AI bro cope; they're just making up whatever nonsense vernacular to make 'stealing and remixing people's art without their permission' sound better.
It's not like there's a government mandated approval process where you have to go to, like, a Department of Arts to get a seal of approval to be able to purchase a license of Zbrush, but ONLY if you're one of Elon Musk's kids and the last letter of your first name is '8'.
No, you literally google 'zbrush' click the first link and buy the sub or whatever, then make art.
Democratization of tools (or skill) would be to provide access to those tools/knowledge and the economic security to practice that skill to underserved demographics who otherwise can't afford it.
This is not what ML is doing. That's just a bullshit argument and techbros know it. But at this point most things they say are bullshit (AGI right around the corner, anyone? 😂), so that's not surprising.
What they're doing is "socializing" the labour of an already vulnerable to economic downturns class, artists. You study, practice for years, build a portfolio, and they scrape it to avail themselves of the all labour that went into it, by force and without compensation. It's a parasitic approach that goes beyond copyright infringement or not: as an artist you're the one creating the data being used (otherwise they could just use blank pages, right?), the foundation of their business, you're just having no choice or compensation for the trouble. You're working for them for free. For staunch capitalists they sure love some corporate welfare.
The end result is further concentration of capital in the hands of obscenely rich corporations and their boards at the expense of workers. But this isn't as good soundbite as "democratization of tools", and it's easy to try to ride the wave of AI applications that try to actually help workers (eg better denoisers) instead of admitting gen AI is anything but a way further to exploit then replace them.
There are properly applications for AI which are helpful; like finding some possible molecule for a better medicine; or some star constellation in some very far far galaxy ( which might be questioed by some people ); or some other very scientific and/or compute intensive applications; in fact some new rendering techniques (not complete generating of "content". Now "they" start to replace the middle management managers (the ones sitting between the chairs of the workers and the decision making financial managers). Exaggerating and not really meant too pesimistic but a little bit of sarcasm : Just wait when someone starts to replace the high money making managers.. Finally when "everyone" is replaced by some AI to make "cheaper" products who will be able to buy this ?? The "rich ones" might not even by "cheap" products..
Something came up in a thread recently, the use of an AI image generator for providing art feedback, and I'm curious to hear people's opinions on this.
I'm personally torn, as I see @d1ver 's point about embracing tools that help make us better artists, and which allow him to provide more feedback than he would otherwise be able to afford the time to do. This community is all about providing feedback, his commentary has been solid, and he's a long-time member.
However, calling it a paintover seems like it could be a bit misleading. Maybe it's ok to use "paintover" as a general term for altering an artwork to provide feedback? Looks to me like no Photoshop-esque painting was done, and it perhaps distorts the prevalent understanding on these forums of what a paintover is. Perhaps charitably though you could say inpainting is a form of paintover?
As this thread demonstrates, some AI image generators are infringing on copyrights and on livelihoods. Yet they are without a doubt very powerful tools that artists themselves are using to improve their work. Curious what people think about the use of these tools here in Polycount threads.
Also https://imaginemore.art was shared as a way to inspect art. It's a pretty cool site, worth a look, and has some great tools for art inspection and inspiration. Some of the art shown seems to be public domain (18th century paintings) while other pieces are likely under copyright (stills from modern movies). I wonder if this falls under fair use. There doesn't seem to be a TOS or any About page to explain things, or how uploaded images are used. Though he did provide some clarifications here and here.
I'll ping @MBauer17 and @Noren since this thread seems like a good place to debate these topics. As long as we keep things calm and respectful, I think a debate can be helpful.
Personally I've found that best use case for generative AI is using chatGPT as a search engine tool if google isn't turning up anything for me. If I need really obscure information, and just don't know how to word it or search for it. I'm making an MMO, and have no idea how to setup servers and anticheat and key programmer language, but I'll certainly be asking AI how to get started before doing a deep dive, so that I know how and where to start.
AIgen, particularly for use in feedback, while well intentioned, is risky, and ultimately a bit lazy IMO. Feedback should be teaching people about the fundamentals; how and why an image looks the way it does, why it looks good, so that they have the knowledge to create what they want and quickly identify what's wrong.
The issue with all generative AI is that it does not understand context. The generator here added a really harsh lighting to the scene that I don't think compliments stylistic painterly art at all. d1ver is correct in talking about fore, mid, and background, but in the generated image the camera angle is completely different, putting the tower in a valley instead of up on a hill. The perspective and angle of the roofing is parallel with the ground, instead of angled up to create the sense of fresh air, adventure, and freedom in the alps. The image went from "Oh I'm playing WoW! Happy, fun times!" to "I'm slaughtering monsters as Geralt." Which is also pretty happy fun times, but a different kind for sure.
As some added feedback for MBauer17, you can think about negative and positive space as well, and possibly having less background and more tower (so crop the sides more, scoot the trees in), as it's the focus of the image. Depth of field would also be a good thing to add (or increase). The flowers in the generator is a good idea, but it's completely out of focus and a blurry mess that detracts from where you want the audience to focus on; the tower. The smoke trail is also purposefully artistic, I'd expand even more on that, maybe look into World of Warcraft's Mists of Pandaria artwork for inspiration there. If I don't have time to give feedback, the least I could do is send them to someone who can; I always thought fengzhu school had a lot of really good tutorials for concept work and teaching basics like this. https://www.youtube.com/@FZDSCHOOL/videos
As someone who always struggles through art with "this looks vaguely wrong and I don't know where to start fixing" sort of problems, I'd think the critique and paintovers are valuable for their precision.
Even hallucinations aside, won't an AI "paintover" risk carrying too much visual information that might distract from the point the critic tried to make?
"which allow him to provide more feedback than he would otherwise be able to afford the time to do."
That probably touches on the the crux of the whole AI problem. It is a tool. However we feel about it, it's very likely here to stay. Quality aside, it enables us to do more with less people which in some cases means fewer jobs. It helps with some mundane tasks but can also do the steps we enjoy the most. Reiterating the obvious, I guess.
As for the paintovers, I don't care how they are called, but I'd like to see a clear disclaimer that AI was used. E.g. "AI Overpaint" might downplay the human factor too much, but just add a sentence that Generative AI was used in the process. Like said earlier, it doesn't change the feedback, but it does make a difference how long someone just spent helping someone on an interpersonal level.
Coming back to the quality aspect, AI is trained on illustrations that tend to have good composition and artistic choices, so while some of it is a jumbled mess, it tends to have better composition and "choices" than what a beginner artist could do. So it could be a learning tool, especially if the changes are explained, but there's also a disheartening side to this. "So you spent days and weeks on this, but here is a much better, polished looking version that AI did in 5 minutes." To be clear, I'm not necessarily talking about @d1ver or any of the artworks he provided feedback for in this case. Like mentioned, it's clear there was substantial intervention from his side (at least that's my assumption) and the artworks he started from weren't the kind of beginner work I was referring to. But to be frank, I also don't want this to become a "hey, I ran this through AI for you".
Then there is the question of the exact process. AI can be processed locally, but often enough it means upload of the images to a third party for processing, and ideally, that would be something that we don't simply decide for others. Obviously, there are web crawlers that at the very least create a hash of an image anyway, but I feel like this is a step further.
Lastly, I don't believe there are any truly "ethically sourced" sets of AI training data out there, at least when it comes to bigger ones. E.g., I don't think many people were aware they are signing up for this kind of thing when releasing their artworks under a permissive CC license.
Yeah, I don't want to get into the business of giving feedback to feedback, but this is part of a meta-discussion and other than the mountain range in the background, I didn't really love this one. The different framing, lighting and the flowers felt a bit formulaic and too much removed from the original, and I think the close direct comparison to a specific different scene was detrimental in this case. In my opinion, it was simply impressive for the sheer amount of work that seemed to have gone into it at first sight and its polished looking nature.
A positive example (in my book) would be the Monastery Refuge thread, which also highlights the point of enabling people to give more feedback, as that thread would likely have gone without any before d1ver's reply.
So I had a long heart-felt talk with my personal therapist (AKA my husband) about this. When it comes to AI I feel like I am a single drop of water trying to fight back a blazing fire; I know AI is here and I'm not really thrilled about it. In a sense, I am very much like the human "computers" in the 1950s calculating math equations watching their jobs getting replaced by machines. In this metaphor, I have two options: I can try to fight this inferno and evaporate into history or I can figure out how to benefit the most out of this new technology.
I strongly believe I cannot win this fight. I see the opportunity that AI provides particularly in receiving feedback but only when it's taken contextually because let's be honest, throwing a lens flair and a HDR filter on an image isn't a paintover, let alone a valid criticism to take home and work on.
However, @d1ver had some very good ideas that I can use - primarily he emphasized adding a foreground and bringing up the background mountain range. This is a fundamental in art that I missed completely and would have not taken to heart had he not mentioned it. He quickly went over some great notes that, lets be honest, normally require a lot of time and energy that many posts are unfortunately not given enough attention to. Just look at all the unresponded Final threads and you'll see what I mean.
Am I going to take this as a learning opportunity? Yes absolutely. I got some great feedback among some not so great. As I mentioned to @d1ver in the post, AI is something that requires nudging and evaluation and his career history is what helps him guide him through this critiques.
Onto the other subject at hand, I am not exactly thrilled to know that he is using this community as a platform to not only advertise his AI algorithm, but actively feeding it artists' work without their permission. He is freely going through posts and scraping up WIP and final stuff to use for his own (potentially) monetary gain or future monetary gain. That's not cool.
My ideas for resolution:
State at the front of critique post "This is an AI's interpretation of improvement" or something along those lines.
Prohibit/limit the use of image data scraping (I don't know how to do this, I wish I was smart enough to.)
I came to Polycount seven years ago when I quit my old career to go into this one. It was the place where the oldheads shared their ancestral vertex knowledge with the next generation. I respect and value the insight those members provide and I still respect it. Please keep it that way. Don't let this become the next Pinterest.
Thanks for the post, and I'm glad to see Polycount has been great for you!
To be fair, there's a lot of self-advertising going on at Polycount. It's kind of inherent in any community of professionals and hobbyists. So I personally don't have any issue with people promoting their projects/portfolios/softwares. I think it's a great way for the community to learn about new things, and for members to get the word out about their passion projects.
I think your criticism here is fundamentally about the use of AI image synthesis. Normally I would say it's just another image editing tool, and thus you could use it without mentioning so. However given the longstanding controversies around unethical data scraping, and the strong pushback from artist communities like ours, I think it makes sense for people to note if AI tools are used in the content they post.
I think it's unfair to suggest d1ver is "actively feeding it artists' work without their permission ... freely going through posts and scraping up WIP and final stuff to use for his own (potentially) monetary gain or future monetary gain" when his replies have made it clear he is expressly not doing this.
Also, longtime members here definitely get the benefit of the doubt, presumed innocent until proven guilty, etc. I guess it depends on how the tools he has used are working, and whether uploaded images are used for further training or not. I would love to hear more details from him about the tools he is using, and how the data is handled.
Thanks for surfacing this, @Eric Chadwick - I think it’s an important conversation, and it's great to speak more about it.
Just to clarify: if I understand correctly, the main issue is the perceived time it takes to produce a paintover? Or perhaps some folks feeling misled or cheated if it took less time than they expected?
Maybe it's just me, but I never saw the feedback here as a way to impress anyone. This industry is full of people who deeply need thoughtful, honest critique — I was one of them. And the support I received here changed my life.
Polycount exists to offer that kind of feedback. I'm here to get helpful information across. If we start valuing how impressive someone appears doing it over how useful their insight is, then I don’t believe we’ve earned the right to say that we care about artists.
Oh and thank you for the shoutout for https://imaginemore.art/ We've not officially launched it yet, hence a few things missing but I'm happy to add the TOS and add a thread about it in case there are any more questions. I'm quite excited about how we get to connect artists to the best of the art fundamentals 🤍
I don’t begrudge artists who personally chose to use AI as a source of feedback on their work but I don’t think it belongs on this site.
If artists want AI feedback it’s not hard to go and find it, there are any number of competing models already out there. Polycount has always been a place to get human feedback on human art. I don’t want to see it devolve into a sea of bots and puppet accounts spamming AI feedback to drive engagement with their models.
I don’t mind @d1ver promoting his own service but leaving low effort feedback on people’s threads is not the way to do it. Artists have a reasonable expectation to consent before their work is fed to an algorithm. As has been said, calling it a “paintover” is misleading, it’s twisting the phrase to obscure that it’s 100% an AI render.
if I understand correctly, the main issue is the perceived time it takes to produce a paintover? Or perhaps some folks feeling misled or cheated if it took less time than they expected?
No, rather it's the use of AI/ML image manipulation tools.
@Eric Chadwick What is your opinion on feedback itself that's AI generated? I don't know if d1ver is or is not writing up his own feedback or pushing it through a chatGPT program. Aside from the use of AI to render an image, what concerns should we have if the words of feedback are themselves pushed through an AI program?
Normally I would say it's just another image editing tool, and thus you could use it without mentioning so. However given the longstanding controversies around unethical data scraping, and the strong pushback from artist communities like ours, I think it makes sense for people to note if AI tools are used in the content they post.
That's my feeling on the use of AI in posts here. Personally I dislike most AI-generated content. I find it mushy inconsistent and often tacky. But I do see the benefit of incorporating it into thoughtfully-curated content, it can help speed up the process of generating new content, but this requires a heavy hand in editing.
I'm absolutely against using AI without any human input, and I'm absolutely against using AI models that scrape their training data indiscriminately without regard to copyright. The courts are slow to catch up to the bleeding edge here, but the immorality of many of these models is pretty obvious at this point.
Onto the other subject at hand, I am not exactly thrilled to know that he is using this community as a platform to not only advertise his AI algorithm, but actively feeding it artists' work without their permission. He is freely going through posts and scraping up WIP and final stuff to use for his own (potentially) monetary gain or future monetary gain. That's not cool.
Thanks for your post, @MBauer17. I hear the frustration and uncertainty you're expressing. The emotional tension is real, and I think a lot of us are still trying to find our footing around this.
While I understand your concern I have to reiterate strongly that accusing someone of stealing your data or using this community unethically, without evidence, is exceptionally serious. You are entitled to your worry, but this unsubstantiated slander is no way acceptable. I'm open to a real conversation about what it means to use AI responsibly in creative spaces. But if it's not grounded in facts or mutual respect I can' t see how this is valuable for this community.
Thanks @d1ver what tools did you use to make your feedback for the watch tower? I think it's important to be completely transparent when using AI image manipulation tools.
Edit: It would go a long ways towards addressing the suspicion that artists' work might have been fed into an AI system that was trained using unethically-sourced data. How was the AI model trained? What data was fed into the system?
BTW love the links you posted, going to be watching these tonight!
@Eric Chadwick The exact text feedback comes from me. The image model under the hood is not something that is available to general public, it runs custom on my own infrastructure. It is built not to generate new images, but to interpret feedback and provide tweaks and modifications within the scope of an existing image. Strictly a feedback tool that can't drift far from the original or provided feedback. No artist adverse middle men. All data storage policies are managed by me.
OK thanks, and sorry about the stealth edit in my post. But what data was your model fed on?
There is no feedback loop and the model is not retrained in any way so all the data that comes through us gets immediately destroyed after use. The data sources for initial training are as ethical as it's realistic for any company to establish at this point. On the frontend fair use is clearly defined as transformative, educational and non-competitive with the original. And so our main focus is on making sure that it can't be used for anything other than steering your existing artwork 20% in some direction. It won't create new images and it won't make bad ones magically good. It explicitly follows direction, so it can't think for you. Bad feedback in or bad art in - bad art out. It you don't know art fundamentals - it's not going to help you. Fixing it strictly within the fair use bounds on the usage side.
Well, the problem with "AI feedback" isn't about the pictural quality of the result - as, just like with anything AI, the output of art generators is objectively very convincing now.
The fundamental issue is wether or not one chooses to participate in skill atrophy. Not just for oneself, but also for the community of artists as whole. If everyone leverages "AI feedback", then sure we'll see tons of pretty portfolios ... but they'll also all look very similar (even moreso than currently) and no one will actually learn anything in the process besides an ability to blindly execute what the plagiarism machine dictates.
Now that said I personally have no delusion about it : many, many people will absolutely use this sort of stuff, because as a beginner (especially in a tech-adjacent field like CG art) it's easy to miss the fact that learning (and sweating on it quite a bit) is actually the fun part, and the appreciation for this grind and what comes out of it comes with time. But I personally feel sad for anyone relying on this stuff even in the slightest, because they'll miss out on a lot of art history, on tracking back the influences of masters and the masters before them, and most importantly on connections with fellow artists. It's absolute cultural apathy just for the sake of "catching up" and "making a pretty image", and it robs the artist of any ownership on their production. This is not something I would personally wish on anyone.
There are probably some nuances to the above depending on how said "AI feedback" is developped, but at the end of the day, regardless of how ethical or clever it is, it will always be lumped into the category of "AI-enhanced" hence putting the artist in the "can't quite do a finished project without leveraging AI" category ... even though we have literally thousands of years worth of examples of how to do things with our simple, limited meat brains.
Replies
I've personally been enjoying Chromeo a whole lot for years now. Of course they mix everything digitally and are highly technical (they even have a very interesting video dissecting one of their tracks) but a certain analog feel still seeps through, and apparently they manage to perform great live sets for a 2 men band.
(Never head This Masquerade, this is great
https://www.linkedin.com/posts/sixte-de-vauplane_ai-machinelearning-genai-activity-7262747642663776257-j2DY?utm_source=share&utm_medium=member_desktop
The tool will be exclusive to Pocoyo and any licenses Animaj acquires or partners with, though they will release some tools to the public.
There are many problems the with the entertainment industry as it stands now and I feel it's difficult and fuzzy to point to any one area and shout THERE"S THE PROBLEM! I'll attempt to share a few thoughts that are loosely connected with the hope of forming more coherent thoughts on the matter.
The "sameness" was a trend long before we got to AI, in fact I think it may be an inescapable principle of scaling in general. I saw a research study in the early 2010s that showed that if you take portraits of several groups of people from around the world and split them into different sets of images, then average the sets of images together the result is that each set produces roughly the same face no matter the widely ranging individual looks. I felt this hinted at a larger principle as it described what I was starting to see emerging in the games industry. Fast forward to 2024, While we have a very wide array of indie games, anything AAA has trended towards Pixar if it's stylized or towards hyper realism if it's something like COD. I think AI accelerated this 10x as that seems to be what it's good at. (accelerating what's already there) I really do feel like overall most things in production or recently reduced can be described as "soulless". My kids watched the Mario movie and all I could think was, this is fun but it has no soul. We're in a time of transition and the role of the artist won't end with us, but it will change radically. Many of the things going on today need to change, and I think we all feel it. I don't think all the lawsuits will stop the progress. If we look at the automotive factory workers from Detroit I think we can all see where this goes.
We still have automotive factory workers, even at Tesla. But the job has changed drastically. I'm not sure I will like the new world but I'm genuinely not sure what to do about it.
I do think we will see small shops pop up with all human crafted material at a premium but that will be a very competitive market with only a few winners.
Sounds like a scam. These AI aren't content aware at all, let alone on the level you'd need to be to understand a storyboard and transfer that into actual 3D movement, especially filling in all the stages inbetween. Doesn't convince me when, without signing in, 90% of the comments are people who call themselves "angel investor" or "generative AI developer" or such. And I mean...If I wanted animation, I'd go for an animator.
Thanks for posting that. I have to say this is a great use of AI. There's still a bit of AI jank happening on that face. Though it's much less uncanny than the early days of CGI characters and as they say "This is the worst it will ever be." I get the feeling that this AI tooling is moving in the direction of being locked away in abstraction for the vast majority of game developers with Nvidia and other big companies offering the masses a turn it on or off option when using it to develop a game. Maybe that's good, maybe that's bad and maybe I'm wrong. Just thinking out loud here.
The other shown technologies are much more interesting imho.
Yeah. How do you like having yr stuff copied for profit, you prick! (scrapers)
https://sourcingjournal.com/topics/technology/shopify-artificial-intelligence-tobi-lutke-flexport-klarna-1234743438/
They tend to be very tech-forward, but I'm not sold on AI actually replacing whole persons. It can work great as a tool, but it needs constant guidance from a human to correct errors.
Great example recently on LinkedIn, how AI-generated code can be really problematic. https://www.linkedin.com/posts/michael-kisilenko-ceo_softwareengineering-codequality-technicaldebt-activity-7313843508874743808-GkXN
https://blender-mcp.com/
https://x.com/sidahuj/status/1899460492999184534
Its from a team based in Delhi India that developed this so its good that it brings more visibility to startups in India
Users are reporting that they aren't getting consistent and accurate results, but it seems useful for prototyping.
This seems to be the best result of its use, but not sure if it can be replicated,
https://x.com/sidahuj/status/1905196751125987629
I mean that end result is nothing like the input, besides yeah some toony fox and racoon characters. Can anyone even claim to be in control here, art directing with such a gamble is... Kinda pointless
Unless it all just does not matter
This one seemed a bit different in its approach, though probably comes down to how people use it.
The usual gacha AI approach has a market, just not sure its something that can be sustainable when it comes to monetization.
Here's one promoter, I used to give him feedback on linkedin but he blocked me in the end.
https://www.tiktok.com/@zk_lonelion/video/7374769029522787590?lang=en
https://www.theguardian.com/technology/2025/jun/09/apple-artificial-intelligence-ai-study-collapse
These are generalizations , I understand things are way more nuanced than this:
No need for 3D Modeling = no need for Maya, Max, Blender, C4D and so on.
No need for rendering / texturing = no need for Vray, Redshift, Substance etc
Of course these companies are doing their AI work to incorporate it into their software and "stay relevant". But I honestly just don't see it happening. At least not in the coming 10 years? AI slop is that, slop, it's boring, it always has "that look", and it's just tacky.
I sometimes compare it to the "not everyone should have a podcast" thing. Just because you can does not mean you have to or that you will create quality things.
The democratization of tools is there, and that's awesome, but just because you can now "make an animation with a prompt" does not mean that
a) it's going to be good
b) people will actually consume it let alone PAY for it.
I don't know, I'm just confused at this moment being a 3D Artist, I'm keeping myself busy, working and learning but holy crap the AI ghost creeps in frequently and I hate the feeling.
It's not like there's a government mandated approval process where you have to go to, like, a Department of Arts to get a seal of approval to be able to purchase a license of Zbrush, but ONLY if you're one of Elon Musk's kids and the last letter of your first name is '8'.
This is not what ML is doing. That's just a bullshit argument and techbros know it. But at this point most things they say are bullshit (AGI right around the corner, anyone? 😂), so that's not surprising.
What they're doing is "socializing" the labour of an already vulnerable to economic downturns class, artists. You study, practice for years, build a portfolio, and they scrape it to avail themselves of the all labour that went into it, by force and without compensation. It's a parasitic approach that goes beyond copyright infringement or not: as an artist you're the one creating the data being used (otherwise they could just use blank pages, right?), the foundation of their business, you're just having no choice or compensation for the trouble. You're working for them for free. For staunch capitalists they sure love some corporate welfare.
The end result is further concentration of capital in the hands of obscenely rich corporations and their boards at the expense of workers. But this isn't as good soundbite as "democratization of tools", and it's easy to try to ride the wave of AI applications that try to actually help workers (eg better denoisers) instead of admitting gen AI is anything but a way further to exploit then replace them.
Exaggerating and not really meant too pesimistic but a little bit of sarcasm
Just wait when someone starts to replace the high money making managers.. Finally when "everyone" is replaced by some AI to make "cheaper" products who will be able to buy this ?? The "rich ones" might not even by "cheap" products..
This.. seams to be interesting:
https://slashdot.org/story/25/06/21/1755225/what-if-customers-started-saying-no-to-ai
or even that:
https://entertainment.slashdot.org/story/25/06/22/1747250/how-the-music-industry-is-building-the-tech-to-hunt-down-ai-generated-songs
..sadly this does not really help those who are already layed of..
I'm personally torn, as I see @d1ver 's point about embracing tools that help make us better artists, and which allow him to provide more feedback than he would otherwise be able to afford the time to do. This community is all about providing feedback, his commentary has been solid, and he's a long-time member.
However, calling it a paintover seems like it could be a bit misleading. Maybe it's ok to use "paintover" as a general term for altering an artwork to provide feedback? Looks to me like no Photoshop-esque painting was done, and it perhaps distorts the prevalent understanding on these forums of what a paintover is. Perhaps charitably though you could say inpainting is a form of paintover?
As this thread demonstrates, some AI image generators are infringing on copyrights and on livelihoods. Yet they are without a doubt very powerful tools that artists themselves are using to improve their work. Curious what people think about the use of these tools here in Polycount threads.
Also https://imaginemore.art was shared as a way to inspect art. It's a pretty cool site, worth a look, and has some great tools for art inspection and inspiration. Some of the art shown seems to be public domain (18th century paintings) while other pieces are likely under copyright (stills from modern movies). I wonder if this falls under fair use. There doesn't seem to be a TOS or any About page to explain things, or how uploaded images are used. Though he did provide some clarifications here and here.
I'll ping @MBauer17 and @Noren since this thread seems like a good place to debate these topics. As long as we keep things calm and respectful, I think a debate can be helpful.
AIgen, particularly for use in feedback, while well intentioned, is risky, and ultimately a bit lazy IMO. Feedback should be teaching people about the fundamentals; how and why an image looks the way it does, why it looks good, so that they have the knowledge to create what they want and quickly identify what's wrong.
As some added feedback for MBauer17, you can think about negative and positive space as well, and possibly having less background and more tower (so crop the sides more, scoot the trees in), as it's the focus of the image. Depth of field would also be a good thing to add (or increase). The flowers in the generator is a good idea, but it's completely out of focus and a blurry mess that detracts from where you want the audience to focus on; the tower. The smoke trail is also purposefully artistic, I'd expand even more on that, maybe look into World of Warcraft's Mists of Pandaria artwork for inspiration there. If I don't have time to give feedback, the least I could do is send them to someone who can; I always thought fengzhu school had a lot of really good tutorials for concept work and teaching basics like this. https://www.youtube.com/@FZDSCHOOL/videos
Dave Rapoza is also pretty awesome, he does 2d art but art fundamentals are basically 1:1 transferable to 3d. https://www.youtube.com/@DaveRapoza/streams
https://www.youtube.com/@JamesGurney
https://www.youtube.com/@StevenZapataArt
https://www.youtube.com/@abcwithsamwise-ABC
"which allow him to provide more feedback than he would otherwise be able to afford the time to do."
That probably touches on the the crux of the whole AI problem. It is a tool. However we feel about it, it's very likely here to stay. Quality aside, it enables us to do more with less people which in some cases means fewer jobs. It helps with some mundane tasks but can also do the steps we enjoy the most. Reiterating the obvious, I guess.
As for the paintovers, I don't care how they are called, but I'd like to see a clear disclaimer that AI was used. E.g. "AI Overpaint" might downplay the human factor too much, but just add a sentence that Generative AI was used in the process. Like said earlier, it doesn't change the feedback, but it does make a difference how long someone just spent helping someone on an interpersonal level.
Coming back to the quality aspect, AI is trained on illustrations that tend to have good composition and artistic choices, so while some of it is a jumbled mess, it tends to have better composition and "choices" than what a beginner artist could do. So it could be a learning tool, especially if the changes are explained, but there's also a disheartening side to this. "So you spent days and weeks on this, but here is a much better, polished looking version that AI did in 5 minutes."
To be clear, I'm not necessarily talking about @d1ver or any of the artworks he provided feedback for in this case. Like mentioned, it's clear there was substantial intervention from his side (at least that's my assumption) and the artworks he started from weren't the kind of beginner work I was referring to. But to be frank, I also don't want this to become a "hey, I ran this through AI for you".
Then there is the question of the exact process. AI can be processed locally, but often enough it means upload of the images to a third party for processing, and ideally, that would be something that we don't simply decide for others. Obviously, there are web crawlers that at the very least create a hash of an image anyway, but I feel like this is a step further.
Lastly, I don't believe there are any truly "ethically sourced" sets of AI training data out there, at least when it comes to bigger ones. E.g., I don't think many people were aware they are signing up for this kind of thing when releasing their artworks under a permissive CC license.
Yeah, I don't want to get into the business of giving feedback to feedback, but this is part of a meta-discussion and other than the mountain range in the background, I didn't really love this one. The different framing, lighting and the flowers felt a bit formulaic and too much removed from the original, and I think the close direct comparison to a specific different scene was detrimental in this case.
In my opinion, it was simply impressive for the sheer amount of work that seemed to have gone into it at first sight and its polished looking nature.
A positive example (in my book) would be the Monastery Refuge thread, which also highlights the point of enabling people to give more feedback, as that thread would likely have gone without any before d1ver's reply.
Edit: added a direct link.
To be fair, there's a lot of self-advertising going on at Polycount. It's kind of inherent in any community of professionals and hobbyists. So I personally don't have any issue with people promoting their projects/portfolios/softwares. I think it's a great way for the community to learn about new things, and for members to get the word out about their passion projects.
I think your criticism here is fundamentally about the use of AI image synthesis. Normally I would say it's just another image editing tool, and thus you could use it without mentioning so. However given the longstanding controversies around unethical data scraping, and the strong pushback from artist communities like ours, I think it makes sense for people to note if AI tools are used in the content they post.
I think it's unfair to suggest d1ver is "actively feeding it artists' work without their permission ... freely going through posts and scraping up WIP and final stuff to use for his own (potentially) monetary gain or future monetary gain" when his replies have made it clear he is expressly not doing this.
Also, longtime members here definitely get the benefit of the doubt, presumed innocent until proven guilty, etc. I guess it depends on how the tools he has used are working, and whether uploaded images are used for further training or not. I would love to hear more details from him about the tools he is using, and how the data is handled.
Just to clarify: if I understand correctly, the main issue is the perceived time it takes to produce a paintover? Or perhaps some folks feeling misled or cheated if it took less time than they expected?
Maybe it's just me, but I never saw the feedback here as a way to impress anyone. This industry is full of people who deeply need thoughtful, honest critique — I was one of them. And the support I received here changed my life.
Polycount exists to offer that kind of feedback. I'm here to get helpful information across. If we start valuing how impressive someone appears doing it over how useful their insight is, then I don’t believe we’ve earned the right to say that we care about artists.
We've not officially launched it yet, hence a few things missing but I'm happy to add the TOS and add a thread about it in case there are any more questions. I'm quite excited about how we get to connect artists to the best of the art fundamentals 🤍
@Eric Chadwick
I don’t begrudge artists who personally chose to use AI as a source of feedback on their work but I don’t think it belongs on this site.
If artists want AI feedback it’s not hard to go and find it, there are any number of competing models already out there. Polycount has always been a place to get human feedback on human art. I don’t want to see it devolve into a sea of bots and puppet accounts spamming AI feedback to drive engagement with their models.
I don’t mind @d1ver promoting his own service but leaving low effort feedback on people’s threads is not the way to do it. Artists have a reasonable expectation to consent before their work is fed to an algorithm. As has been said, calling it a “paintover” is misleading, it’s twisting the phrase to obscure that it’s 100% an AI render.
I'm absolutely against using AI without any human input, and I'm absolutely against using AI models that scrape their training data indiscriminately without regard to copyright. The courts are slow to catch up to the bleeding edge here, but the immorality of many of these models is pretty obvious at this point.
Thanks for your post, @MBauer17. I hear the frustration and uncertainty you're expressing. The emotional tension is real, and I think a lot of us are still trying to find our footing around this.
While I understand your concern I have to reiterate strongly that accusing someone of stealing your data or using this community unethically, without evidence, is exceptionally serious. You are entitled to your worry, but this unsubstantiated slander is no way acceptable.
I'm open to a real conversation about what it means to use AI responsibly in creative spaces. But if it's not grounded in facts or mutual respect I can' t see how this is valuable for this community.
And to repeat myself again - no scraping or training on anyone’s work has ever occurred, period. No artist data was being stored, harvested, or used anywhere. Insinuating that I'm any less of an artist than you or give any less of a shit about this is exceptionally unfair and shortsighted. I will put my record standing up for artists rights against anyone here any day.
https://gdcvault.com/play/1034178/ARTificial-Ethics-and-Aesthetics-of
https://gdcvault.com/play/1029014/Creative-Future-with-Artificial
And if you want to have a genuine conversation around our positive contributions towards mitigating the risk of AI for artists - I'm here.
Edit: It would go a long ways towards addressing the suspicion that artists' work might have been fed into an AI system that was trained using unethically-sourced data. How was the AI model trained? What data was fed into the system?
BTW love the links you posted, going to be watching these tonight!
The exact text feedback comes from me. The image model under the hood is not something that is available to general public, it runs custom on my own infrastructure. It is built not to generate new images, but to interpret feedback and provide tweaks and modifications within the scope of an existing image. Strictly a feedback tool that can't drift far from the original or provided feedback. No artist adverse middle men. All data storage policies are managed by me.
The data sources for initial training are as ethical as it's realistic for any company to establish at this point. On the frontend fair use is clearly defined as transformative, educational and non-competitive with the original. And so our main focus is on making sure that it can't be used for anything other than steering your existing artwork 20% in some direction. It won't create new images and it won't make bad ones magically good. It explicitly follows direction, so it can't think for you. Bad feedback in or bad art in - bad art out. It you don't know art fundamentals - it's not going to help you. Fixing it strictly within the fair use bounds on the usage side.
There are probably some nuances to the above depending on how said "AI feedback" is developped, but at the end of the day, regardless of how ethical or clever it is, it will always be lumped into the category of "AI-enhanced" hence putting the artist in the "can't quite do a finished project without leveraging AI" category ... even though we have literally thousands of years worth of examples of how to do things with our simple, limited meat brains.