🛠 MOSS-Transcribe-Diarize 0.9B: Comprehensive Audio Processing
The open-source MOSS-Transcribe-Diarize 0.9B model has been released, combining speech-to-text (STT) transcription and diarization (speaker separation) into a single pass. This allows for obtaining structured text with timestamps and speaker labels without using two separate systems. The model supports over 50 languages and recordings up to 90 minutes in duration.
🌍 The transition to end-to-end architectures reduces latency and errors. Its compact size (0.9B parameters) makes the solution accessible for local deployment.
👤 It is now possible to create high-quality subtitles and meeting minutes locally, using one small model instead of heavy cloud APIs.
Source 1: https://github.com/OpenMOSS/MOSS-Transcribe-Diarize Source 2: https://huggingface.co/OpenMOSS-Team/MOSS-Transcribe-Diarize