VerbatimDrop a recording
Any audio or video file: mp3, m4a, wav, flac, ogg, or a video with an audio track. Drag it onto the window, use Open, or right-click a file in Finder and choose Verbatim.
Transcription that never leaves your Mac.
Drop in a recording and get clean, searchable text with speaker names, ready for Word, Markdown or subtitles. Every step runs on your own Mac. No cloud, no account, no subscription.
Download on the Mac App Store
Coming soon · Free 7-day trial, then $8.99 once
Runs 100% on-device
No account, no analytics, no ads
Works offline from the first launch
Pay once, use forever
Demo
From recording to transcript in one drop.
A five-minute meeting goes in, a labelled transcript comes out, and nothing touches the network. Twenty-four seconds, no sound.
Recorded on an M-series Mac with the built-in Base model
Speaker names and formatting on, both running locally
How it works
Three steps, zero uploads.
Verbatim bundles the same open-source speech models that power research labs, compiled to run on Apple silicon. Your audio is decoded, listened to and written up without a single byte going to a server.
Your Mac does the listening
Whisper transcribes the speech. A second model works out who is talking. A small language model fixes punctuation, capitalisation and paragraphs, and gives speakers their names when they are mentioned.
Search, copy, export
Find any phrase, copy the whole thing, or export to plain text, Markdown, a Word document, or SRT and WebVTT subtitles with timestamps.
Features
Built for interviews, meetings and lectures.
Speaker names, not just labels
Verbatim marks every change of speaker. When people address each other by name in the recording, it uses those names instead of "Speaker 1" and "Speaker 2". You can also tell it how many people were in the room.
Watch it happen
Text appears while the recording is still being processed, with a live progress estimate. A one-hour meeting takes a few minutes on a recent Mac.
- Choose from five Whisper models, from fast to most accurate
- Automatic language detection, or pick one of twelve languages
- Cancel at any time
Export where your notes live
Plain text, Markdown with headings and paragraphs, a formatted Word document with bold speaker names, or SRT and WebVTT subtitles. Timestamps are optional everywhere except subtitles.
Your settings, your models
Everything that runs is listed in Settings, with its size, and can be deleted at any time. The speech model ships inside the app. AI formatting and speaker detection are optional downloads, each asked for explicitly before anything is fetched.
Privacy
Private by construction, not by policy.
Cloud transcription services keep your audio on their servers, often to train their models. Verbatim has no server. The app is sandboxed by macOS, makes no network requests on its own, and only contacts the internet when you ask it to download an optional model. There is nothing to opt out of because nothing is collected.
- Speech recognition
- Whisper via whisper.cpp, Metal-accelerated. The Base model (142 MB) is built in; Tiny, Small, Medium and Large-v3-turbo are optional downloads.
- Speaker detection
- pyannote community models running in Core ML (22 MB, optional). Requires macOS 15.
- AI formatting
- Gemma 3 4B via llama.cpp (2.5 GB, optional). Every passage it rewrites is checked against the raw transcript; your words are never changed.
- Input
- mp3, m4a, aac, wav, aiff, flac, ogg, mp4, mov and anything else macOS can decode. Audio is resampled to 16 kHz mono in memory.
- Output
- .txt, .md, .docx, .srt, .vtt, or the clipboard.
- Requirements
- macOS 14 Sonoma or later on a Mac with Apple silicon. About 1 GB of free disk for the app, more if you add models.
- Network
- Only for model downloads from huggingface.co, started by you. No telemetry, no crash reporting, no update pings.
- Data kept
- Your preferences, and the models you chose to download. Transcripts exist only where you save them.
Pricing
Try it for a week. Then pay once.
The 7-day trial unlocks everything. After that, one purchase through the Mac App Store unlocks Verbatim permanently on every Mac signed in to your Apple Account. Because the models run on your hardware, there is no monthly bill to keep it going, no server that can be switched off, and no per-minute charge for long recordings.
| Transcribing 10 hours a month | Over one year | Verbatim |
|---|---|---|
| Typical cloud transcription plan, $10 to $30 a month | $120 to $360 | $8.99, once |
| Pay-per-minute API, about $0.006 a minute | $43 plus your own setup | $8.99, once |
| Where your audio goes | Their servers | Stays on your Mac |
One-time purchase
$8.99once, no subscription
- Free 7-day trial with every feature, no card required
- All five speech models, speaker detection and AI formatting
- Unlimited recordings, unlimited length
- Works on every Mac you own, through the App Store
- Future updates included
FAQ
Questions people ask.
Does it really work without an internet connection?
Yes. The speech model ships inside the app, so you can transcribe on a plane. The optional speaker-detection and formatting models are downloaded once, when you turn those features on, and then work offline too.
How accurate is it?
It uses OpenAI's Whisper models, the same family behind most transcription services. The built-in Base model is fast and good for clear speech. For accents, background noise or technical vocabulary, switch to the Medium or Large-v3-turbo model in Settings.
Which languages are supported?
Whisper understands about a hundred languages. Verbatim detects the language automatically, or you can set English, Spanish, French, German, Italian, Portuguese, Dutch, Japanese, Chinese, Korean or Russian explicitly for better results.
Does it change what people said?
No. AI formatting only adds punctuation, capitalisation and paragraph breaks. Every passage the formatter touches is compared with the raw transcript, and any wording change is reverted to the original words.
What happens when the trial ends?
The app stays installed and shows the purchase screen. Buy the full version to keep transcribing. If you already bought it on another Mac, choose Restore Purchases.
Can it record meetings or use the microphone?
Not yet. Verbatim works from files. Record with Voice Memos, QuickTime or your meeting app, then drop the file in.
Which Macs can run it?
Any Mac with Apple silicon (M1 or later) running macOS 14 Sonoma or newer. Speaker detection needs macOS 15. Intel Macs are not supported.
Your recordings. Your Mac. Your words.
Free for 7 days, then yours for good.
Download on the Mac App Store