BaoCut ships local moss-transcribe-diarize model with speaker ID, alongside qwen3-asr and whisper

huangyun_122 · x · 2026-10-29

A user spotted that BaoCut, the video editor recommended by dotey, has built in Fudan's moss-transcribe-diarize model (1.73 GB, runs locally) behind a one-toggle "identify speakers" feature.

It's listed as the recommended model alongside other options: local qwen3-asr-0.6b (10x realtime, high accuracy), whisper-large-v3-turbo (8x realtime), and the cloud-based, per-minute billed gpt-4o-mini-transcribe — letting users switch between local and cloud transcription freely.

Related event: BaoCut integrates Fudan's moss transcription model with local speaker diarization(2 posts)→

Original post →

More from Apps

Apps channel →