Hi everyone,
Here's a workflow that let you rotate a character in 360 deg using WAN 2.2 TI2V unified model.
Be sure to pass the prompt to chatGPT while uploading an image of your character and tell it to update the prompt according to your character (or do it yourself lol)
Description
-- First version --
FAQ
Comments (10)
I've tried it but I keep getting OOM errors. Does this work on 16GB VRAM?
It should work. Did you try adding a "Block swapping" node ?
@YouYouTheBoxx Thx, that fixed my problem. Another question, is it normal for the character to turn into an asian real person at the end? No matter what I try I can't get the same results as you.
@reborndrako Don't use context frames and limit your generations to 121 frames at most.
@YouYouTheBoxx That was it. Awesome results now! Thx so much for sharing this amazing workflow!
@reborndrako Could you specify what all you changed in the workflow, to get it to work for you? Also, did you use the same model the workflow selects, or your own variation of Wan 2.2?
@verdurewoks34352 I used the workflow without changing anything except bypassing the WanVideo Context Options node and lowering the frames from 161 to 121.
https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models
first one i could find of the model.
Wan Video Sampler node show: "Unsupported CUDA architecture: sm75" mean Unsupport 20xx series and older cards, can it be run compatibility by modifying parameters
If you have sage attention: it doesn't work for 20xx unless using 1.X (If I recall correctly).