End-to-end streaming speech-to-speech simultaneous interpretation engine (Hy-SimulMT-style reference): audio chunking + VAD, SimulMT read/write policies, cascade ASR->MT->TTS and direct S2S, voice cloning, StreamLAAL latency, BLEU/GEMBA/WER eval. Pure Python, zero dependencies.
python realtime tts zero-dependency asr voice-cloning speech-translation speech-to-speech simultaneous-interpretation simulmt streamlaal
-
Updated
Oct 27, 2026 - Python