What it does
Generate videos from text or an input image using released model weights. The official repository includes ComfyUI guidance and image-to-video generation instructions.
A useful first task
Animate one product shot
Start from one still image and request a simple, short movement. Compare the result frame by frame.
How to start
- Follow the official repository or its ComfyUI guide to download and run the model locally. The documented baseline uses Linux and an NVIDIA CUDA GPU.
- Start from one still image and request a simple, short movement. Compare the result frame by frame.
- Check the result before relying on it.
Where it falls short
The documented baseline minimum is 14 GB of GPU memory with model offloading. Installation and settings take work; a reported local trial describes visual artifacts.
What remains your job
Choose the usable take and edit the final sequence.