3 ms·
The end-to-end speech-to-speech claim is interesting, especially avoiding the ASR→LLM→TTS pipeline, which is where most latency and error compounding happens.
by kuandriy 9mo ago
The end-to-end speech-to-speech claim is interesting, especially avoiding the ASR→LLM→TTS pipeline, which is where most latency and error compounding happens.