Feature Request: Add Real-time Voice Chat Demo

[freddyaboulton]

Overview

It would be great to have a demo that enables natural voice conversations with omni-mini-2 using WebRTC technology. This would allow users to interact with the model through speech in real-time, both locally and in cloud environments.

Existing Implementation Reference

I already have a working implementation for omni-mini-1 on Hugging Face Spaces that demonstrates this capability:

https://github.com/user-attachments/assets/040ec236-b336-4a14-880a-f9cde82e63e1

Key Benefits

• Real-time voice interactions
• Works both locally and in cloud/remote environments
• Enhanced user experience through natural conversation

Technical Resources

• Reference implementation for omni-mini-1: Source Code
• WebRTC component documentation: gradio-webrtc

Potential Implementation

The port from omni-mini-1 to omni-mini-2 should be relatively straightforward given the existing implementation. This could be a valuable addition to enhance the project's interactive capabilities.

Would this be a feature the maintainers are interested in adding to the project?

[superFilicos]

thank you!

[mini-omni]

@freddyaboulton hi, your demo looks really great. You are very welcome to contribute, and I would be happy to test it together. Thank you.