>>281740510
And as AI image generation improves, this will become more and more of a viable and common workflow for things like town scenes.
Have an anime that largely takes place in the same town? Just model the town once in 3D. Then throw renders of the town into AI gen with edge detection.
The end result will look better than just slapping on some image filters over a render, as has been done a lot in the past, and will be faster than manually tracing/shading over a render.
Image gen is improving faster than LLMs are - LLMs seem to be plateauing a bit, actually. I don't think it will be long before the above workflow produces results that you won't be able to tell are AI unless you pause the anime and really study the still frame.