My first WAN LoRATraining data was made with my paperfold image LoRA <a target="_blank" rel="ugc" href="https://civarchive.com/models/1152352/paperfold-style-make-folded-paper-objects-animals-and-humans">https://civarchive.com/models/1152352/paperfold-style-make-folded-paper-objects-animals-and-humans</a>v0.6 The trigger is "paperfold" and prompts like a "a paperfold cat" is enough to get the style, and to get animation use long descriptive prompts like this: "A determined female paperfold triathlete runs toward the finish line with intense focus. As she crosses it, the camera slowly zooms in capturing her triumphant moment. She raises her arms in a victorious pose, smiling with pride and exhaustion. The crowd cheers in the background, and the sun casts a warm glow over the scene, emphasizing her achievement."Awaiting the first NSFW animation :Dv0.6 training data contained some videos bu mostly images and does animation with lower strength (around 0.7).v0.5 The trigger is "p4p3rf0ld" but you also need to use "made of folded paper" for it to work at normal strength. I will train a better version but this is working with a little luck and good prompting.v0.5 struggles with animation and is only trained on images I can't get comfyui to save metadata so some prompts are from memory

Folded paper style - WAN2.1 t2v

Better captioning on this version and also videos including in the training data

Files with SHA256: 81184153f9ab37800248ec4eb8d5ab99b020ad6e3bd68500a2a3ac5beae9daeb