xiaopengde

Murmur converts Chinese or any Whisper-supported meeting and interview recordings into text using local Whisper large-v3, then cleans the output into markdown with speaker labels, corrected ASR typos, and chapter divisions, with optional docx export. It does not support real-time transcription or heavy multi-speaker overlap.

It targets users who need readable transcripts directly in VS Code or Word, teams replacing paid services like Feishu Minutes, Tongyi Tingwu, or Otter.ai, and developers running end-to-end transcription inside coding assistants such as Claude Code, Copilot, Codex, or Cursor.

Murmur is cross-platform: it uses mlx-whisper on macOS with Apple Silicon and whisper-ctranslate2 on Windows, Linux, and Intel Macs. All processing happens locally with zero cloud calls and zero subscriptions, so recordings never leave the machine.

Yes, murmur is fully local with zero cloud dependency and zero subscription fees, so audio privacy never leaves your device. It serves as a free replacement for paid transcription services like Otter.ai, Feishu Minutes, and Tongyi Tingwu.

Murmur requires a local Whisper large-v3 runtime via mlx-whisper or whisper-ctranslate2. It is unsuitable for real-time transcription, noisy meetings with heavy speaker overlap, or scenarios demanding strict speaker diarization accuracy.