MOSS Releases Multi-Speaker Transcription Model

pmttyji · reddit · 2026-07-09

OpenMOSS-Team released MOSS-Transcribe-Diarize 0.9B on Hugging Face. This end-to-end audio understanding model performs multi-speaker transcription, speaker diarization, timestamping, and acoustic event identification for long audio or video. It generates structured transcripts with anonymized speaker labels in a single pass, targeting scenarios like meetings, podcasts, interviews, and classrooms.

Related event: MOSS Releases Open-Source 0.9B Audio Transcription Model Supported by vLLM(5 posts)→

Original post →

More from Multimodal

Multimodal channel →