Hi ,
We are using the Gemini Live Translate API for real-time speech-to-speech audio pipelines. In high-load backend environments, managed runtimes (Python/Node.js) introduce garbage collection pauses and single-thread bottlenecks that hinder low-latency audio streaming.
To meet our performance needs, we built a working native C++ client using Boost (Asio/Beast) for low-latency PCM audio streaming. It has proven very efficient, and we are happy to open-source or contribute our C++ implementation to the community as a reference if helpful.
Long-term, maintaining custom WebSocket protocol layers carries maintenance risks. We would love to know if an official, production-ready (GA) C/C++ SDK is on the roadmap to support high-concurrency WebRTC gateways and real-time media servers.
Thanks for the great work on Gemini!