ComfyUI_OmnimatteZero

OmnimatteZero

Official implementation of OmnimatteZero: Training-Free Video Matting and Compositing via Latent Diffusion Models

Update

• 改成comfyUI的patcher封装/ change to ComfyUI patcher encapsulation

1. Installation

In the './ComfyUI /custom_nodes' directory, run the following:

git clone https://github.com/smthemex/ComfyUI_OmnimatteZero.git

---

2. Models

• Vae a-r-r-o-w/LTX-Video-0.9.7-diffusers
• Dit or gguf or int8 Chinese user aliyun or huggingface smthem/LTX-Video-0.9.7-diffusers-merge

--  ComfyUI/models/vae
    |-- LTX-Video-0.9.7-vae-diffusers.safetensors #vae rename
--  ComfyUI/models/diffusion_models # optional 
    |-- LTX-Video-0.9.7-comfy-int8_convrot.safetensors #  LTX-Video-0.9.7-diffusers.safetensors
--  ComfyUI/models/gguf  # optional 
    |-- LTX-Video-0.9.7-diffusers-Q8_0.gguf # need gguf nodes

3. Example

4. Citation

@inproceedings{samuel2025omnimattezero,
  author    = {Dvir Samuel and Matan Levy and Nir Darshan and Gal Chechik and Rami Ben-Ari},
  title     = {OmnimatteZero: Fast Training-free Omnimatte with Pre-trained Video Diffusion Models},
  booktitle = {SIGGRAPH Asia 2025 Conference Papers},
  year      = {2025}
}
``