A universal voice translator for Windows that works with any application - Zoom, Discord, games, browsers. Not a plugin or extension, but an OS-level solution with under 300 ms end-to-end latency.
Two parallel pipelines provide bidirectional translation: your speech is translated and routed to a virtual microphone for the other person, while their audio is captured from system output, translated, and played back through your speakers. Any language pair is supported.
Built with streaming speech recognition, LLM-powered translation with caching, neural text-to-speech, voice activity detection, and low-level Windows audio integration. 1500+ lines of production code.
Stack: Python, Streaming STT, LLM Translation, Neural TTS, VAD, WASAPI, Virtual Audio Driver
Two parallel pipelines provide bidirectional translation: your speech is translated and routed to a virtual microphone for the other person, while their audio is captured from system output, translated, and played back through your speakers. Any language pair is supported.
Built with streaming speech recognition, LLM-powered translation with caching, neural text-to-speech, voice activity detection, and low-level Windows audio integration. 1500+ lines of production code.
Stack: Python, Streaming STT, LLM Translation, Neural TTS, VAD, WASAPI, Virtual Audio Driver