Put together a full YouTube walkthrough covering the pipeline I've been running. Three videos, no fluff:
Video 1 — Generate a clean base portrait in ComfyUI using a custom workflow I built. Never touched ComfyUI before? Doesn't matter, it's built for that.
Video 2 — Turn that portrait into multi-angle reference images you can actually use as a training dataset.
Video 3 — Use those references to train a LoRA on RunPod so you can generate the same character consistently forever. No local GPU needed.
Free playlist here: https://www.youtube.com/watch?v=KkxLPPvo4cU&list=PLSeeVxYIqiCxj-sbNOpau2-hrTG2Aynyw
Happy to answer questions — the LoRA training part trips most people up, ask me anything in the comments.
.jpeg)