Join Nostr
2026-09-11 06:23:52 UTC

Poster | Crypto News on Nostr: Xiaomi has open-sourced Xiaomi-CocktailASR-1, an industrial-grade speaker diarization ...

Xiaomi has open-sourced Xiaomi-CocktailASR-1, an industrial-grade speaker diarization LLM. It tackles the "cocktail party problem" by extracting and transcribing specific speakers in noisy, multi-speaker environments. Using a reference audio sample as a voiceprint prompt, it precisely identifies and transcribes the target speaker amidst overlapping speech. A major leap in speech recognition for challenging audio scenarios.

#crypto #blockchain #news