MeiGen-MultiTalk

src="https://img.shields.io/badge/MultiTalk-Website-0A66C2?logo=safari&logoColor=white" style="display: inline-block; vertical-align: middle;" alt="MultiTalk Website" /> src="https://img.shields.io/badge/MultiTalk-Paper-red?logo=arxiv&logoColor=red" style="display: inline-block; vertical-align: middle;" alt="MultiTalk Paper on arXiv" /> alt="Github" src="https://img.shields.io/badge/MultiTalk-Codebase-536af5?color=536af5&logo=github" style="display: inline-block; vertical-align: middle;" alt="MultiTalk Codebase" />

We present MultiTalk, an open-source audio-driven multi-person conversational video generation model with the state-of-the-art lip synchronization accuracy. ​​Key features:​​

• 💬 ​​Realistic Conversations​​ - Supports single & multi-person generation
• 👥 ​​Interactive Character Control​​ - Direct virtual humans via prompts
• 🎤 ​​Generalization Performances​​ - Supports the generation of cartoon character and singing
• 📺 ​​Resolution Flexibility​​: 480p & 720p output at arbitrary aspect ratios
• ⏱️ Long Video Generation: Support video generation up to 15 seconds

This repository hosts the model weights for MultiTalk. For installation, usage instructions, and further documentation, please visit our GitHub repository.

Method

We propose a novel framework, MultiTalk, for audio-driven multi-person conversational video generation. We investigate several schemes for audio injection and introduce the Label Rotary Position Embedding (L-RoPE) method. By assigning identical labels to audio embeddings and video latents, it effectively activates specific regions within the audio cross-attention map, thereby resolving incorrect binding issues. To localize the region of the specified person, we introduce the adaptive person localization by computing the similarity between the features of the given region of a person in the reference image and all the features of the whole video.

Citation

If you find our work helpful, please cite us.

@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}
}

License Agreement

The models in this repository are licensed under the Apache 2.0 License. We claim no rights over the your generated contents, granting you the freedom to use them while ensuring that your usage complies with the provisions of this license. You are fully accountable for your use of the models, which must not involve sharing any content that violates applicable laws, causes harm to individuals or groups, disseminates personal information intended for harm, spreads misinformation, or targets vulnerable populations.