ComfyUI PyramidFlow Wrapper
Wrapper for PyramidFlow -models: a/https://github.com/jy0205/Pyramid-Flow
Nodes (8)
A full decode for people who don't have time
The node that actually makes the video
The dual-clip brain behind your prompt
Free speed if your torch cooperates
The checkpoint dropdown and the fp8 trap
Turning latents into frames without OOMing
Your starting image's trip into latent space
The quiet node your img2vid can't skip
ComfyUI wrapper nodes for Pyramid-Flow
UPDATE
As the first Flux version is out, I'm dropping the SD3 support and refactored the whole thing, if you still want to use the old nodes they are archived in the legacy branch
The fluxmini version can run with 7GB VRAM, currently it only supports 5 second videos (temp 16). Fp8 severely reduces quality and is not recommended, only use it if you must.
Download models from:
https://huggingface.co/Kijai/pyramid-flow-comfy/tree/main
To ComfyUI/models/diffusion_models and ComfyUI/models/vae
https://github.com/user-attachments/assets/1372549a-4b4e-4569-a062-8f72880e8c4e
https://github.com/user-attachments/assets/d0bd38eb-6378-4cfa-ae55-1b4498b7ce84
Original repo: https://github.com/jy0205/Pyramid-Flow