Frequently asked questions

No. Your recordings are read from disk and transcribed by the processor in your Mac or iPhone. No audio and no transcript is ever sent anywhere.

The app does make network connections, and it is worth being precise about what they are: it talks to RevenueCat to check whether you have Pro and to fetch current prices. That is the only thing in the app that uses the network. There is no analytics on what you transcribe.

You do not have to take our word for any of this. Turn on Airplane Mode and transcribe a file — it still works, because the model is inside the app. Our guide on checking whether an app is really offline covers two more thorough tests.

Accuracy depends far more on your recording than on any setting. Clean audio — one person, a decent microphone, a quiet room — transcribes very well. Accuracy drops with background noise, people talking over each other, distance from the microphone, strong accents, and unusual names or technical terms.

Rather than trusting a number, test it. The app is free to download and transcribing is free and unlimited, so you can run your own worst recording through it and read the first few segments before paying anything. That tells you more about your audio than any published benchmark.

If your recordings are consistently difficult, our guide on Whisper model sizes explains when a larger model genuinely helps and when it does not.

Free lets you transcribe as many files as you like, as often as you like. There is no minute allowance and no file limit. You can read the first five segments of each transcript.

Pro unlocks the full transcript and turns on exporting — TXT, VTT, SRT and CSV, for one file or for the whole queue at once.

The split is deliberate: transcription is the expensive part and it is free, so you can check the app handles your audio before deciding whether it is worth paying for.

You can choose. Pro is $2.99 a month, $7.99 a year, or $12.99 once for a lifetime licence with no renewal.

Most people should take the lifetime option. It costs less than five months of the monthly plan and roughly a year and a half of the yearly one, and it never bills again.

Purchases are handled by the App Store, so subscriptions are managed and cancelled in your Apple account settings, and refunds go through Apple rather than us.

Considerably faster than the recording itself on Apple silicon — a one-hour interview typically finishes in a few minutes. The exact time depends on your chip, the model, and what else the machine is doing.

Because everything runs locally there is no upload wait and no queue behind other customers, which in practice often makes it faster end to end than a cloud service, even one running on better hardware.

Queue a folder of files and hit Transcribe once. The app works through them in order while you do something else.

No. There is no per-file length limit, no monthly minute allowance, and no cap on how many files you transcribe.

This is the practical difference between local and cloud transcription. A cloud service meters you because every minute costs it server time — Otter includes 1,200 minutes a month on its Pro plan, Descript counts uploaded hours. Nothing is being metered here because the work happens on hardware you already own.

Very long recordings take proportionally longer and use more memory, but they are not blocked.

Audio in: mp3, wav, m4a, ogg, aac, caf Video in: mov, mp4 — the audio track is used, so you can drop a video file straight in Out: TXT for plain notes, VTT and SRT for subtitles, CSV when you want the timings in a spreadsheet

If your file is in another format, convert it first. Nothing needs converting for the supported ones — drop them in as they are.

99 languages, from Afrikaans to Yoruba. The full list is on the home page and you can search it.

The language can be detected automatically or set by hand for each file. Setting it explicitly is usually more reliable when a recording is short or noisy.

Accuracy varies by language. Whisper's training data is heavily weighted towards English, so English transcribes best and the less widely spoken languages are noticeably weaker. Test a real recording in your language before committing.

No. You get one continuous transcript with timestamps, not labelled speakers.

For a one-on-one interview this is usually easy enough to fix by hand afterwards. For a panel discussion or a large meeting, it is real work, and an app with speaker diarization will save you hours — our MacWhisper comparison covers that honestly.

It runs on any Mac with macOS 13.0 or later, Intel included.

It is much faster on Apple silicon, where the work runs on the Neural Engine. On an Intel Mac transcription is meaningfully slower, though still perfectly usable for occasional files.

The app is free, so if you are on Intel the honest advice is to download it and time a real recording before buying Pro.

Yes. The model ships inside the app, so there is nothing to download at runtime and no server to reach.

The only thing that needs a connection is buying or restoring Pro, since that goes through the App Store. Once Pro is active, transcribing and exporting work with the Wi-Fi off indefinitely.

No. It is a single app for macOS and iOS, so one purchase covers both. Buy Pro on your Mac and it is active on your iPhone and iPad under the same Apple Account, and the other way round.

If it does not appear on a second device, open the app there and use Restore Purchases.

Not at the moment. The app produces the transcript and exports it; corrections are made in whatever you open the file with afterwards.

For subtitles this is rarely a problem, since an SRT is plain text and opens in any editor. Our guide on making an SRT file covers reading it through for name errors before you import it into Final Cut or Premiere.