# MeiGen-MultiTalk Demo

This is a demo of [MeiGen-MultiTalk](https://huggingface.co/MeiGen-AI/MeiGen-MultiTalk), an audio-driven multi-person conversational video generation model.

## Features

- 💬 Generate videos of people talking from still images and audio
- 👥 Support for both single-person and multi-person conversations
- 🎯 High-quality lip synchronization
- 📺 Support for 480p and 720p resolution
- ⏱️ Generate videos up to 15 seconds long

## How to Use

1. Upload a reference image (photo of person(s) who will be speaking)
2. Upload one or more audio files:
   - For single person: Upload one audio file
   - For conversation: Upload multiple audio files (one per person)
3. Enter a prompt describing the desired video
4. Adjust generation parameters if needed:
   - Resolution: Video quality (480p or 720p)
   - Audio CFG: Controls strength of audio influence
   - Guidance Scale: Controls adherence to prompt
   - Random Seed: For reproducible results
   - Max Duration: Video length in seconds
5. Click "Generate Video" and wait for the result

## Tips

- Use clear, front-facing photos for best results
- Ensure good audio quality without background noise
- Keep prompts clear and specific
- For multi-person videos, ensure the reference image shows all speakers clearly

## Limitations

- Generation can take several minutes
- Maximum video duration is 15 seconds
- Best results with clear, well-lit reference images
- Audio should be clear and without background noise

## Credits

This demo uses the MeiGen-MultiTalk model created by MeiGen-AI. If you use this in your work, please cite:

```bibtex
@article{kong2025let,
  title={Let Them Talk: Audio-Driven Multi-Person Conversational Video Generation},
  author={Kong, Zhe and Gao, Feng and Zhang, Yong and Kang, Zhuoliang and Wei, Xiaoming and Cai, Xunliang and Chen, Guanying and Luo, Wenhan},
  journal={arXiv preprint arXiv:2505.22647},
  year={2025}
}