ComfyUI-OpenDiTWrapper
Wrapper nodes for OpenDiT: a/OpenDiT, supports Open-Sora t2i and i2i
Nodes (7)
OpenSora's text brain is bigger than its diffusion model
The node that drags OpenSora v3's diffusion model onto your GPU
The VAE that understands time — not the one you already have
Your prompt gets dressed up before OpenSora ever reads it
The sampler where OpenSora actually makes the video
The decoder that turns video latents into actual frames
The node that turns image-to-video from a dream into a wire
WORK IN PROGRESS
https://github.com/kijai/ComfyUI-OpenDiTWrapper/assets/40791699/1cb3aa06-1576-43ad-a01e-f5a63f59127d
https://github.com/kijai/ComfyUI-OpenDiTWrapper/assets/40791699/422565ae-aea9-43a7-b9d5-a8a8e7963141
Tested to work in Linux, and on Windows with current ComfyUI portable (pytorch ):
pytorch version: 2.3.1+cu121
xformers version: 0.0.26.post1
VRAM requirements are pretty hefty, but far less than the original repo due to offloading. For example 48 frames at 768x512 fit in 15GB, less frames/smaller res can fit 10GB.
Installing:
pip install -r requirements.txt
pip install xformers --no-deps
for portable (in ComfyUI_windows_portable -folder):
python_embeded\python.exe -m pip install -r .\ComfyUI\custom_nodes\ComfyUI-OpenDiTWrapper\requirements.txt
python_embeded\python.exe -m pip install xformers --no-deps
Original repo:
https://github.com/NUS-HPC-AI-Lab/OpenDiT/