Dream Fusion 3D models. Is the end close for 3D artists?
So almost every week, without even looking for it, I see advancements of AI in my feed. It feels like it's getting exponentially better and that the people working on it are drinking coffee nonstop to get this out...
I saw this in my feed yesterday and it seems the end is neared than I thought. Sure, it looks pretty crap so far, but it was the same with 2D art. It feels like in one maybe two years you will be able to make a whole environment that would take months for a team of artists to make in just an hour or so
https://www.youtube.com/watch?v=uM5NPodZZ1U
I am a 3D artist, started this journey almost 5 years ago from nothing, with no artistic background and been doing this professionally for two years. This job gives me my confidence and it would be devastating to me if I got replaced by an unfeeling machine. I can do some concepting and have made some pretty unique hard surface stuff for my client but most of the stuff that I make is made after concept art or an image, exactly the kind of stuff AI can replicate.
Let's say you are making a WW2 game. Why pay an artist to model an STG 44 if you can just tell the AI to model it for you in two seconds? Oh, you are making fictional, highly realistic WW2 weapons? Just feed the AI images of WW2 weapons and tell it to make something similar. Making a sci fi game? Why pay an artist if you can just feed the AI examples of sci fi guns that you like and it will just spit out an excellent result in no time?
I do hard surface and I can't draw. Before this crazy AI stuff got so advanced I figured I could learn how to draw and paint if I ever lost my job but it seems like AI is going to take that away too, at least entry level jobs. It would take me at least three years to get somewhat decent. A year ago the AI couldn't do all the stuff it does now, who knows where it will get in just one or two years? If I have to leave the art world soon, it would take me at least a couple of years to get into another field and if you can replace a high skill job like an artist, you can probably replace other stuff too! I am just a human, and while I think I'm reasonably bright, I'm not a 150 IQ genius or even 130. I can't keep up with this.
Replies
Ah cool. There is the expected !
3D was always a field of rapid evolution. And text to image or text to model does not make traditional art and your 3d skills obsolete. It's just yet another tool in the tool belt. And a great one. As with AI image geneartion, i do not fear of the future of traditional artists. There is still lots of work left. It just shortens the process here and there.
Reminds me a bit to the revolution when sculpting became a thing, and traditional box SDS modeling became obsolete for character creation. Or the PBR workflow. And so on.
We live in exciting times for sure :)
I don't understand. How does it not make my skills obsolete? Again, if you are working on a WW2 game, why hire an artist to model your STG44 or even fictional variants if you can just tell the AI to make it for you? If I decide to go the traditional art route, what entry level jobs are there going to be for me? It's going to take at least a few years until I'm decent and who knows how good this thing will be by then.
PBR and sculpting did not threaten to replace anyone. In fact, they were tools that made the process more artistic. What is artistic about entering a few keywords, then the AI scans Artstation for the best gun models and concepts on Artstation, analyzes their textures, topology and artistic direction and then just spits out 20 fully modeled, textured and animated models? At this rate, in 20 years all we're going to have is regurgitated "art" made by an AI and we're all going to consume the same rehashed content over and over. Maybe the top 5% of 3D artists are going to keep doing what they do and while I am quite good at what I do, I am nowhere near that level.
I don't understand. How does it not make my skills obsolete?
Text to image is gambling, that's why. It's not you want to have a character with this and that features and refine until you have arrived at the goal. You gamble with the right keywords, you gamble with every result. It's throwing a dice. And even the best result needs at least retopo and proper texturing.
PBR and sculpting did not threaten to replace anyone.
But despite the fact i heard the same fears back in the days. I heard it also with VR. Now look what happened to VR. It's evolution. New tools replaces old tools, the workflow changes. But the work remains.
This imho is very different from it just being another tool, and I feel like claims that a human will always be needed don't stand up to scrutiny when the results are objectively of better quality than what most modelers or illustrators can do (either 2d or 3d). Rock solid models derived from 2D modelsheets seem like the next logical step right around the corner. The slightly stylized animal heads in the DreamAI turnarounds are already excellent and far above what most entry-level modelers can do with a sculpting program. A clever AD teamed with a technically minded 3d lead could integrate that very nicely into something great, without the need for much 3D gruntwork. Now whether or not the result will have so-called soul is a different story.
Put differently : an artist will always be needed if there is a desire for a human to generate the bits and pieces in the first place. But there are plenty of cases where absolutely no one cares, just like how no one cares already about their sneakers being put together in a third world sweatshop. As a matter of fact there is already a market for creepy generated videos aimed at kids all revolving around keywords like "colors", "Elsa from Frozen" and "Spiderman" ranking millions of views ; and people are happy to consume the same boring caped hero movies/TV series year after year. These dreadful movie scripts all revolving around some kind of magical glowing cube really aren't much better than something AI-generated ...
All that said, a job is not some kind of favor or reward based on how hard a skill is to acquire - it just means that someone doesn't have the skill or time to do something, and pays someone else to do it instead. Now that run-off-the-mill 2d and 3d content can be generated in seconds, it will be up to people interested in the entertainment field to identify some other/new needs that can be turned into an income. And ironically enough, the ones able to do so are specifically the ones with the highest amount of technical know-how combined with a keen eye for seizing opportunities.
Well, it may make some old jobs obsolete. But new tools have always also lead to new jobs. And the new AI tools are in the wild for a pretty while now. And the graphics world is still not collapsed. Currently it's in the hype phase. Reality will catch it soon enough.
We will see. As told, i see it as an opportunity :)
We will definitely reach that point where most jobs will be obsolete there are plenty of book on transhumanism and futurizm that "predict" that We can only hope that Ai will be supplementary similar to how Scott Eaton is guiding it and AI is doing the heavy lifting and not in some corvair belt sorting way yes no or maybe to what AI has done art director style but in garbage bin
If anyone is actually interested and want to predict the figure doesn't have too look too far just a quick analyse of how DeepBlue and AlphaGo had change their sports will give you are pretty good picture .. in short everyone is playing like a machine
I dont know if anyone had notice but since metahuman and the other similar tool came out there are a lot less ppl making characters and even less interested in anatomy if some blendshape character generator have such impact imagine AI + photogrammetry + houdini like modeling tools or as pior said most of if not all technology has technical orthographic drawing wihich will be very easy to convert to 3d and photogrammetry can be used for organic stuff
The reality is the skill ceiling is moving.
Unreal Nanite "kills" LoD Level.
Midjourney kills Concept Artists.
Quixel kills Texture Designer.
etc. ...
Fuck you Gutenberg that i cant make money for copy a book in half a year.
The concept artists need to step up, maybe Storybord/mood and Picture series less single art pieces for consistency.
The LoD creator must step up to the artist.
Its a race.
the AI thing bothered me from the moment i saw it was being created to generate art. i fully believe in its capability to replicate high quality art instantly, its technology improving exponentially, and it disrupting our industry at its foundation. at the same time, i identify as a type of artist, of mind and heart, that cannot allow myself to fall into a frightened luddite's mentality and trust in observing the world's slow pace of change over time and identify the meanings of its effects.
MOCAP WILL KILL ANIMATION!!!
While it is true that it is indeed "just another tool", I feel like there are massive blind spots in people's appreciation of the consequences of all of this.
Sure, the output of this or that AI algorithm isn't perfect or production ready, and artists are not going to lose jobs overnight because of it. And of course it's neat to have access to such a literal "make art" button - especially for AI-bros who just love the fuzzy feeling of "creating" something by typing in a prompt.
But one of the very real consequences of all this is that we are already at a time where the output of such prompts is directly usable in moodboards and thumbnailing/brainstorming. This is of course a lot of fun to play with (and I would encourage everyone to do so, as this stuff really is fascinating), but that also means that panels like this (100% AI generated) are now a thing :
Now *of course* one could prepare a similar collage with a few hours of googling, so in and of itself that's nothing new and both processes are similarly derivative.
But that also means that inevitably, there will be cases where some artists (the human ones) are going to be tasked to work from something that comes directly from an AI image generation tool, as opposed to coming from a human designer/curator. AI evangelists will of course say that it doesn't matter, that it's just a tool, and that speeding up a process is never a bad thing ; but there will absolutely be some people who will straight up refuse to be part of a process in which the thing that is handed down to them didn't come from the experience, knowledge and craftmanship of a designer (even if they never met that person IRL) but just a text prompt with the always useful "trending on Artstation" bit thrown in for good measure.
This rejection will not be because of the results being bad, but because it simply is less interesting since it means being unable to have any conversions about where an idea came from, what the influences of the art director are, and so on.
So to go back to the example of mocap and animation : sure, mocap didn't kill traditional animation. But will an animator specialized in manual animation (or even just mocap cleanup work) which up until now was always based on human capture, be just as willing to do the same work if based fully on a AI-generated sources ?
I am 100% sure that it won't take long for some studios to experience this first-hand and that we'll see some studios and companies positioning themselves as not using AI in their creation process, regardless of it being a useful tool or not. Because even though no one is owed a job, there is still a very real incentive in making sure that a team is working well together and solving problems and design challenges as a tight unit. Of course the deciding factor will be whether or not it will be sustainable to do so.
I am also very interested to see the reaction of the general public a year or so down the line, when we'll start hearing AI-generated music at the mall, or when we'll start seeing tons of products being advertised as "designed by AI". Some will love it because it's new and interesting, some will consume without giving it any second thought, and some will undoubtedly hate it.
This is not just another tool, the 2d AI was making only blobs a year ago and today is making incredible images in so many different styles, including realistic ones that are already getting very hard to tell they are fake, it's not unthinkable that in 1-2 years from now the AI will be making models with the quality of photogrammetry, why wouldn't it? 3d scan already looks like magic, put a bunch of photos in a software and a 3d model textured comes out.
''it's a gamble" yeah, it's a gable today but this argument implies that it will keep that way forever as if prompting some text for the AI to edit an image or 3d object would be unthinkable in the future.
Also I see a lot of talks about "soul" when AI comes up, then I ask, how many soul was put into all the lattest marvel movies that they pop two a year?, how much soul do that ammo box need?
I wouldn't be worried at all. Look at those models. Bad unusable mushy mesh topology, automatic crappy UVs and blurry textures... artists creating 3D art themselves by hand has the needed ultimate precision, and that need never goes away.
AI taking over art is just motivating me to branch out from art and into programming. Art is already so difficult; if we just push ourselves a little more, learning how to create a full game from scratch in unreal engine is really not a stretch goal. Who needs animators, programmers, etc - when artists these days are expected to know a combination of 5-10 different softwares just to get the job done
There's so many hoops and bullcrud we have to jump through. Poor company management, unpaid art tests, bad salaries. It's so stupid, and I'm honestly tired of the ineptitude of the current state of our industry. Blizzard asked me for an unpaid art test. ****ing Blizzard. The guys who rake in millions from Immortal. The sheer audacity. No wonder they have an employee strike every quarterly.
Additionally, Fromsoft, the company behind Elden Ring apparently pays minimum wage for character artists. Nice one.
Literally we might as well just pump out a AAA game on our own with the skill ceiling so high. Oh, you want to paint, pose, and render your character? Time to learn animation, lighting, and node-based materials. At that point you may as well learn UE5 node-based programming; blueprints. Multiplayer? Making sure your code is consistent and not hackable? Understanding business, accounting, taxes and marketing?
Easy compared to what we've been through already.
There's a bit more involved to create a AAA game than throwing some assets together. As a lone wolf you may be lucky to finish your project at all ^^
But i can definitely recommend to learn some programming. It changes the view at the whole :)
I'm looking forward to the day where I can draw a bounding box and ask the AI to fill it with a placeholder mesh of whatever description.
that just sounds like procgen levels, could probably set that up in a few days with unreal, otherwise a custom blueprint can be made for that, unless you mean something like AI art but in 3d... which is probably something someone is working on already lol
Once people start using Midjourney with google's 3d AI, we're all screwed
https://www.youtube.com/watch?v=H-pTZf1zsa8
@FourtyNights : while the technical characteristics of the models are indeed similar to the output of current 3d scanning tech (automatic meshing + vcols) which indeed is not usable directly in a game if a high visual fidelity is needed, the remarkable thing to me is how the tech does an incredible job at capturing stylized animals better than what most Zbrush artists would. The models have a certain soft focus quality to them (probably as a consequence of the diffusion algorithm) that is very hard to capture in sculpt, especially in Zbrush since the software has always been very poor at giving accurate visual feedback on models.
Put differently : even though the outcome is still visually less punchy than what a good team can do (Art Director with a strong visual sense + talented modelers), the results are IMHO already aesthetically superior to what most 3d modelers would do on their own when asked to create a stylized lion or frog without a perfectly crafted model sheet to work from.
If anything, another unexpected side effect of this tech is how it highlights the obvious limitations that our traditional CG art software has. By that I mean : in a day when an AI can generate good images and models in under 30 seconds, the convoluted and counter-intuitive processes we've grown to accept suddenly become ridiculous - like how 2d artists have been faking spontaneity and brush strokes by placing scans of IRL acrylic paint streaks over their digital paintings after the fact, or how people have come to accept the incredibly counter-intuitive Zbrush interface as the norm. Now all of a sudden having to navigate countless menus and keyboard shortcuts to do something as seemingly simple as a beveled edge becomes more absurd than ever.
hey look, it's already here https://twitter.com/JunGao33210520/status/1573310606320484352
I am skeptical. Perhaps those AI things could replace some simple routine but videogames have so many challenges having not a simple solution and nobody knows what's right . Things requiring a taste to nuances and deep down access to control.
And so far AI in image model generation looks same as AI support in my bank. They laid all people off and now you can't communicate with the bank without lots of bad swearing.
So far I see AI is able to generate mostly useless visual information noise. Not a single AI based tool that is really helpful in the pipeline. Well, helpful but not so much.
Simpligon - only help]s with something simple and typical. Still better to do lods by hand.
Adobe Sampler - inconvenient as hell , only does anything good with small pebbles on the ground , nothing else really . Other Ai text-to-texture generators are just ridiculous.
Photoshop AI selection - select nicely only people and kittens , other AI tools are just toys and nothing really helpful in sight. I'll probably stop my subscription soon. Nothing new and really helpful there for a decade already.
Ai models and UV - a polygonal and UV mess. Not a chance for something efficient for typical game render engine.
I've been subscribed to what was Unity Art engine for a year , that only Ai thing that was remotely useful . Nevertheless I used Substance Designer and Photoshop patch tool instead. Art engine never did what you wanted . Always something weird. Not a sense of how repeating the material is coming. No way to really control anything. Same as communicating with AI support in my bank.
We hear about this cool AI stuff for a decade already and nothing really helpful yet. Instead it's getting hard to find anything good in Internet because a huge flood of senseless Ai stuff , a huge wave of random visual noise is coming.
You all should realise that one day games are done by AI
reviewed by AI
bought on a pile of shame by AI
and played by AI
biological human race is doomed
Happy Armageddon!
The good thing is , humans have then time for the really important things again :)
Traditional art live on !
We'll see if AI art is as successful as elephant and chimpanzees art was :)
Everyone currently excited about or embracing procedural generation should be even more excited about this because it's more intelligent procedural generation.
If you're worried that a robot will take your job, make yourself more useful :p
"Work expands so as to fill the time available for its completion." - Parkinson's Law
I've been dabbling with Stable Diffusion for a few weeks now and the more I use it, the less worried I am for artists.
As someone who manages people in my professional life, I simply do not have the time to be typing prompts into things and evaluating results. The context switching alone isn't worth it.
Yes, AI's can iterate very quickly, but a hundred iterations on an idea isn't useful. I need ten at most, and ideally they should be the best of the best of the original hundred.
From those ten, I would probably want combinations of five of them, as well as miscellaneous cleanup and obviously elaboration from the artist.
From there, we'd probably develop the concept as time allows.
Art is very much about standing out from the pack, especially when you want to sell a product. If the pack all have AI's, the quality level is still going to average out and artists will still be needed to bring the quality level up another notch.
Though I'd love to feel confident with just adopting these new tools to our work environment, unfortunately we generally evaluate AI "artistic" capabilities on our own experiences with Stable Diffusion, Midjourney, Dall-e or something else we can get access to. Yet it is extremely short time to evaluate anything, not mentioning completely new technology, and these generators are neither the only ones nor the last ones to appear in the nearest future. I guess we don't know learning pace of neural networks models, so we can not correctly guess anything on our today's experience basis.
Secondly, their founders admit that, yes, in the first few years there will be just modifications in artististic roles and even new jobs titles will appear, but soon after that workplaces for creatives will decrease and it is inevitable.
Thirdly, there are other AI tools in development that are already better in delivering accurate concepts and more usable 3D models (I believe some Nvidia tools and Google Parti are going in that first direction; Nvidia MoMa concentrates on generating 3D objects for professional modelers).
We don't have the knowledge about the next steps of technological companies and there's so much going on. E.g., how many other labs are working right now on more efficient AI models and what Chinese companies are doing? What if AI labs will colaborate with each other on even bigger scale, how that will fasten development? How it will grow with fresh money from new investors, lured by this crazy hype everywhere? And besides giants, what about "atomistic" activity from smaller dev teams around the world? Perhaps with open source solutions they'll boost AI growth and availability with their own apps. And so on.
I think it's the stage when artists are troyan horses of AI. Many of us adapt these tools to our work pipelines, introducing them to companies environments. Few years from now they won't be something new and surely will be much more accurate and user friendly... for any user.
Just like every other time something new comes along and the luddites start crying, civilisation won't collapse and instead we'll all end up being able to produce cheaper, better quality stuff to sell to people.
Its absolutely reasonable to posit that a factory produced table leg has less soul than a handcrafted one but you have to also accept it's 100th of the price and much more likely to be the same length as the three that come with it.
"Few years from now they won't be something new and surely will be much more accurate and user friendly... for any user."
25 years ago I had the same hope for usual non-AI tools for content creation: textures , models, FX , etc. it all looked like just around a corner.
25 years later it all is still mad scientists creation you have to waste your life for, monstrously inconvenient and artist unfriendly. Well , just a few minor upgrades basically. A very few really convenient things scattered here in there in different packages.
i forget who said it but i side with it is just another tool, and if they take over we just get more free time for other things.
We are basically organic "robots", you do not force yourself to breathe it does it for you, you do not have to control your heart it just does it. (an algorithm or A.I.)
[spoiler]My art thread will have what i wanted to put so i do not disturb those that would have not "enjoyed" reading the full comment.[/spoiler]
To those artists feeling doubtful about these advances. Just know...
My father who i
sphd in physics told me everyone had been super exited about thermonuclear reactors, tokomacs etc in mid 60's . A new potentially eternal source of energy is just around a corner , a huge flow of grants and dissertations , space traveling prospects etc.60 years after we are still not any closer really, not for a bit and probably never be. Just going to be back at Moon maybe using old good fossil fuel .
Here is a video that perfectly articulates my fear. This not just about art. It's beyond that.
https://www.youtube.com/watch?v=tjSxFAGP9Ss&ab_channel=StevenZapataArt
AI-generated art is replacement for stocks, basically.
It's not really an instrument (that you will be learning for years), it's more like a... condition.
Just using search text keywords there instead of Turbosquid or something.
Lucky here with something as complex as 3D art, look at the latest Adobe keynote, 2D Goalpost has moved very hard in little time.
Choosing and compositing things together will always stay an highly important skill.