Open source voice chat infra that rivals GPT-4o
AI voice chat infrastructure that uses WebSockets. It can achieve voice-to-voice latency as low as 300ms (what GPT-4o does) without a unified voice codec. Everything runs on a single high-end consumer GPU.