Update your comfyUI before using v1.6
This workflow uses minimax viggle-h3 workflow cleanly and quickly for motion& dance, sample preview gens. You can upscale by Minimax H3 Latent Upscaler (3D) + RTX Upscaled for quickly generates. With viggle-h3 lora It doesn't usually have a direct impact, but it's passable. This advanced, two-stage workflow completely solves the budget VRAM crisis.
You must have
- ModelAttentionBackend
- ComfyUI-sol-attn
🛠️ Quick Start
Import the workflow file into your workspace.
Verify all paths point correctly to your quantized checkpoints.
Upload the first reference video for the motion, and a reference picture for the character that will be dancing.
⚡ enjoy
Description
viggle-h3 edited, two-stage workflow completely use Latent Upscaler+RTX Upscaler
FAQ
Comments (5)
Nice thanks, hoping something that can replace scail2 with Minimax... can we use loras on this ?
Hello , thanks for this job.
The results are crazy and fantastics.
What is the real use of "load alpha video " ?
Get the following error without using that option if i use only the main video : "Viggle-Animate Conditioning (H3) 'NoneType' object has no attribute 'shape'".
Should it be possible like a ltx2.3 stefan falkok workflow to have pose tranfert but with another background (with a h3 prompt ? ). This last option should be marvelous !
Have a good day !
You can use the video directly with cond_video (Viggle-Animate Conditioning). As for the prompt, I’ve tried various methods, but we’ll likely have to wait for an update.
I always got the black background, how to fix that?
You can use the video directly with cond_video (Viggle-Animate Conditioning).