I am aware AI is kinda shunned as anti-artist, but I for a while wanted a good human & AI workflow. I don’t want to be beat by AI and I already know I can’t keep up with it in speed, so quality has to be a thing I contribute, but most AI tools output weird meshes, and weird textures. What I want, is AI I can control as a helper for time-saving, and also a quality baseline so I can assure my workflow is never outputting worse than AI quality anywhere.
Then I found Stable Gen add-on for blender and it was pretty cool… but ultimately I was disappointed in that too, but it had enough potential to prove the idea that projecting a camera view onto a model for texturing looks pretty good… until you move the camera. So you need to project from a bunch of cameras.
So i spent days projecting views, tweaking settings, doing human touch ups, and AI generating to cover up my human touch ups to make an asset that’s optimized for game engine use and still looks pretty good.
My first test was a bull:
Of course that was the period of days of mostly experimenting. And yeah, there’s imperfections still, but I at least feel I was able to use AI to a helpful extent while retaining my human artistic freedom while making it.
A good proof of concept but now trying to see If I can make a game ready asset using AI. So I need the workflow more smooth:
Here’s my current project. He’s ~8 hours in un-rigged. 16k face count with a topology easy for (un)subdivision neatly. Still though, my goal is game-engine ready by 8 hours, and he failed to be rigged or animated. Maybe the goal is unrealistic without sacrificing quality or just takes more practice?