05-04-2023, 12:44 PM
Ha, got it about the last layer, thanks! I also understand the advantage of the many layers, it's remotely similar to being able to change the lighting and other aspects of a 3D scene without having to touch the whole model.
The reason why I'm inching toward a one-layer world is mostly because I'm currently going back in time to reconnect to my fundamentals. For the same reason, I'm not interested in generative networks at the moment although from a scientific point of view they are intriguing indeed, and I envision the next phase when the generators manipulate 3D models before rendering.
So for the "behind up" image, did the generator/controlnet create the smoother shading and the highkey-ness of the hair, or did you do a manual pass over it? Pretty convincing result either way!
The examples in your next post are a mix of convincing and creepy especially the more realistic first one which might be difficult to fix. I do see the interest the technique can bring though. Thank you for sharing!
The reason why I'm inching toward a one-layer world is mostly because I'm currently going back in time to reconnect to my fundamentals. For the same reason, I'm not interested in generative networks at the moment although from a scientific point of view they are intriguing indeed, and I envision the next phase when the generators manipulate 3D models before rendering.
So for the "behind up" image, did the generator/controlnet create the smoother shading and the highkey-ness of the hair, or did you do a manual pass over it? Pretty convincing result either way!
The examples in your next post are a mix of convincing and creepy especially the more realistic first one which might be difficult to fix. I do see the interest the technique can bring though. Thank you for sharing!