The Automaton race is baked and in engine. It's the last race I'll create for now, excluding a secret sixth race planned that'll be revealed in a DLC or expansion far in the future. Future plans are a bunch of minor improvements across all races, defect check all races for baking issues, and take these base meshes to the next level by actually painting them in Substance; at the moment they only have Zbrush polypaint vertex colour. Afterwards - basic locomotion animation tree in UE5 & FACS facial animation system.
I'm thinking it's hard to tell whether if you're working for microsoft or nvidia, and what percentage of the trillions invested into generative AI is allocated to marketing
I'm a programmer. My boss always tells me to use AI so that I can write code faster.
Here's how coding with AI works: I tell the AI what to code, and it gives me something that's about 50% of what I described. Then, I tell the AI what's missing. The AI apologizes and provides the remaining 50%. However, the code doesn't run. I tell the AI what the compiler says, and the AI apologizes and gives me a fixed version. I can run it, but it doesn't do what it should. I tell the AI about it. The AI apologizes and gives me an improved version. Now, it's almost doing what it's supposed to do, but it still produces faulty output. I tell the AI, it apologizes, and it fixes the code. The code works as long as the input data is normal, but it fails if the data isn't normal or if it's a rare edge case. I tell the AI that the code isn't robust and doesn't handle edge cases. The AI agrees with me and rewrites the code, but now the code isn't working anymore. I tell the AI that it broke the code. The AI apologizes and fixes the code again. Finally, I have a working piece of code that does what it's supposed to do. However, it looks horrible. It looks as if an amateur wrote it. It's nearly impossible to read. It has poor performance, since it's not optimized at all. If you ever have to add a feature to it or find a bug, God help you. Time spent on this: between an hour or two because all the forth and back and all the test runs in between. Here's how coding without AI works: I write beautiful, fast, easy-to-read code that does what it's supposed to do. It's almost bug-free on the first attempt and considers all edge cases. It just takes a few test runs or a short debug session to find whatever is wrong with it on the first attempt. Time spent on this: About 20 to 40 minutes, depending if tests just run fine or whether I have to do a debug session as well. Why? Because I'm a trained professional who knows his job, and has been writing code for over 25 years. Okay, I hear you say. But even if it took four times longer, a trained professional like you is expensive, and an untrained person could have spent the time with the AI, right? Wrong! An untrained person wouldn't quickly notice that the code produces incorrect output, can't handle invalid data, or ignores important edge cases. I can see those issues at once because I have years of experience and I made those mistakes myself as a beginner. Someone with no programming experience will take that faulty, unstable code, release it, and call it a day. Customers will run away when the app crashes at startup or corrupts data permanently. They'll also have an app whose performance is bad and uses far more memory than required because the code is just poor and only functions minimally. It's like saying, "I don't need an expert. I can repair that gas leak myself," and then having your house explode three days later.
"For me, texturing and sculpting can be especially draining, and sometimes the retopo/UV/baking combo (even with RizomUV or Marmoset) ends up eating way more time than I’d like. It often feels like I’m fighting the process instead of moving forward."
Well, all of this is self-imposed really. Nothing forces you to work on models at a level of detail/fidelity requiring complex pipelines or hundreds of hours.
If the process feels like a chore, the solution isn't to hope for automagical tools (AI or not) to speed it up, but rather, to refocus on something you enjoy more. Or establishing strict time goals and adjusting fidelity and processes around them.
Lastly, wishing for AI to automate the tedious is IMHO a pretty bad idea. Because the day some automagical AI tool will be able to do proper UVs for you, will also be the day when you won't be needed anymore as a modeler because there will be plagiarism machines available for that aswell, and the job of a 3D modeler will only consist of cleaning things up.
If UVs feel like a chore to you know, that's probably just because you are not quite there in terms of modeling efficiently with UVs in mind.
"What do you do if you're trying to make a face and don't have a reference for each angle?"
Well ... one could ask the same thing about every piece of sculpture ever made. The artist simply pulls from their experience, knowledge of anatomy, and personal design vocabulary for the representation (and stylization) of the various facial features.
As a matter of fact you actually don't really want to have references for every angle, because whoever provides these references isn't a perfectly accurate 3D software but rather a mere human artist. The multiple views on the model sheet will never quite perfectly match up, some adjustements will always be required.
A character modeler (or a modeler in general) isn't hired to be a human photogrammetry machine. Having hundreds of hours of drawing, sculpting and modeling under ones belt is an important requirement for the job.
I just finished my rendition of the Sci-Fi Light. I created the light model using Maya then did a minor pass in Zbrush to add a little surface damage and to not make the edges so sharp. I then sent it into Substance Painter for the texturing and finally imported it into Unreal engine for the final render and used Blueprints in Unreal to animate the light.
Vertices: 1801 Tri-count: 3,520 Texture Resolution: 1024 x 1024