Master Transcriber · free for macOS · beta
Master Transcriber turns audio and video interviews into transcripts with the source timecode, with automatic speaker detection, quality control against invented text, translation and clip export. Everything runs on your Mac: no internet connection, no account, nothing uploaded.
Privacy
Your recordings never leave your computer. Once a transcription model is installed, Master Transcriber works on a plane, on location or in a newsroom with no network.
Transcription runs on the Mac's graphics processor with whisper.cpp. Pick your model, from Tiny to Large v3.
Speaker detection runs at the same time on the processor, so the text arrives already split by speaker.
Apple's translation engine translates on this Mac. Each language downloads once, then works offline.
An H.265 720p proxy is made on import. Unplug the drive or card and keep listening, editing and exporting.
Import
Drop audio or video files in the window. Master Transcriber checks the format, counts the audio tracks and reads the source timecode before it transcribes.
Quality control New in 0.3
Whisper sometimes “imagines” text in noise or silence: subtitle credits, sentences repeated in a loop. Master Transcriber stops it in four steps.
Only the passages where someone speaks go to Whisper, in short pieces. Noise and silences are never transcribed.
The language is detected on the voice only, not on the noise at the start of the file.
Repetition loops and typical invented phrases (“Subtitles by…”, “Thanks for watching”…) are taken out.
Words where Whisper hesitates get a dotted orange underline, so you know what to listen to again.
The Quality tab shows the voice detected, the language and the words to check (⌥⌘U for the next one). Removed passages are listed too: listen to them and put one back in a click.
Timecode
The original timecode and frame rate are read from the file (tmcd track, MXF or BWF), from 23.976 to 60 fps, drop-frame or not. That timecode is the reference in the editor, the copies, the exports and the video clips.
Select text and press ⌘C. The quote is copied with its start and end timecodes and the speaker’s name, and with its translation when there is one, ready to paste into your edit notes or script.
[12:25:22:10] Speaker 1 : they are obviously noisy and loud and they are not anybody’s favorite thing to look at… [12:25:45:27]
Translation
Translate the transcript with Apple's translation engine, right on your Mac. The translation sits under each paragraph, smaller and in colour, and follows playback.
Export
Hand the text to a producer, the quotes to an editor, the clip to the edit suite.
Select text and press ⌥⌘E. The segment is copied from the proxy without re-encoding, in seconds, with the right timecode. Tick Burn timecode to print it in the picture, frame-accurate.
Projects and files New in 0.3
See the space each project takes, proxies included, then open, rename, move it to an external drive or delete it.
Rename a file for display, add details (speakers, place, date) and change its start timecode at any time.
Try another model or language: the current text is replaced, but a backup stays in the project folder.
Skipped speaker detection? Run it afterwards: your corrections, highlights and translations are kept.
Your source files are never changed, renamed or deleted, whatever you do in the app.
Models
Whisper models run on your Mac. Install one before your first transcription; after that it works without internet. Everything is in File › Preferences › Transcription models.
| Model | Size | Best for |
|---|---|---|
| Tiny | 77.7 MB | Very fast, limited accuracy: a first look |
| Base | 148 MB | Fast: drafts |
| Small (q5) | 190.1 MB | A good trade-off on modest Macs |
| Medium (q5) | 539.2 MB | Accurate, slower |
| Large v3 Turbo (q5) | 574 MB | Recommended: very accurate and fast |
| Large v3 (q5) | 1.08 GB | Most accurate, slow |
| Large v3 Turbo (full) | 1.62 GB | Full precision, large file |
The q5 versions are compressed: almost as accurate, much lighter. Not sure? Start with Large v3 Turbo (q5).
Keyboard
The shortcuts editors already know, so your hands stay on the keys.
Download
Free. Open the disk image and drag Master Transcriber onto the Applications shortcut. The illustrated PDF manual, in French and English, is in the disk image too.
Download for macOSOn first launch, open File › Preferences › Transcription models and install a model. After that, everything works offline.
If macOS blocks the app the first time, open System Settings, then Privacy & Security, and click Open Anyway.
Details
Built on whisper.cpp, OpenAI’s Whisper models, Silero VAD, FFmpeg, sherpa-onnx, ONNX Runtime, pyannote and WeSpeaker. Full credits and licences are in the app.
Support
Master Transcriber is made by one person. If the app saves you hours of typing, a small tip helps me keep improving it.
☕ Buy me a coffeeContact
Write to me. I read every message.
Master Transcriber