OmnimatteZero

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

Update

• add compose mode，but got a normal effect..
• Test Vram 12G,Ram 64, video 1280x720 5s ,if lower Vram, keep 'block_number' <10 and >0, if 'block_number'=0 ,will run in full mode;
• 小显存block_number 设置1-10，大的可以设置为0 或者大于10，直接调整，distill 模型对于简单的去水印可以试试

1. Installation

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

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

---

2. Requirements

• No need,Perhaps someone may be missing the library.没什么特殊的库,懒得删了

pip install -r requirements.txt

3. Models

• Vae a-r-r-o-w/LTX-Video-0.9.7-diffusers
• Dit or gguf 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-diffusers.safetensors
--  ComfyUI/models/gguf  # optional 
    |-- LTX-Video-0.9.7-diffusers-Q8_0.gguf

4. Example

• compose

5. 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}
}
``