Сonsistency characters | Generate characters only by image and prompt, without character's Lora! | IL\NoobAI Edit Reviews

Hey! I was able to get everything running, thanks again for the great work on this. I did run into a few VRAM-related issues along the way, which surprised me since I’m working with 24GB. (Tesla m40 I can't update pytorch anymore so my comfyui has to be patch for each node it is a super painful process so could be something with not using the most recent pytorch. ) I’m still a little unclear on how the Main Character / Sub Character booleans are intended to function. A brief MD file explaining what each of those toggles does, and the expected combinations at the start of the graph, would be super helpful for new users. It also seems to take quite a long time around 40 minutes per generation. My guess is something on my end is falling back to CPU, but I’m still investigating that. One more suggestion: an MD that highlights where we can choose to bypass pose or depth would be amazing. In complex graphs like this, a few extra comments can really help guide users through the intended workflow especially with so many variable points. I’m also curious about the segmentation approach breaking the character into hands, face, etc. instead of processing the full image at once. Is that primarily for VRAM optimization? Overall, incredible work. I’m using this to generate consistent sprite variations for a game different outfits on the same character and so far it’s working great. If you have any advice on improving final detail (I’m thinking of inpainting after assembling the base sprites), I’d love your thoughts!