MiniMax-Remover is a fast and effective video object remover based on minimax optimization. This Cog wrapper provides a convenient API for running the model on Replicate with video and mask inputs.

You’ll need 2 things to run this model.

Requirements

1. Original video you want to remove object from.
2. Binary Masked video of object you want to remove.

Generate mask video using

1. Get first frame of your original video using https://replicate.com/lucataco/frame-extractor
2. First frame → https://replicate.com/meta/sam-2 → Multiple masked images
3. Select mask of the subject(s) you want to remove.
4. Masked image + Original Video → https://replicate.com/jd7h/xmem → Masked video.

License

This Cog wrapper follows the same license as the original MiniMax-Remover project. See the original repository for license details.

Citation

If you use this model, please cite the original MiniMax-Remover paper:

@article{minimax2024,
  title={MiniMax-Remover: Taming Bad Noise Helps Video Object Removal},
  author={Bojia Zi and Weixuan Peng and Xianbiao Qi and Jianan Wang and Shihao Zhao and Rong Xiao and Kam-Fai Wong},
  year={2024}
}