You talk, it types, in whatever app you are already in
On the Mac it lives behind a global shortcut. Press it, say the sentence, and the words arrive at the cursor. In Mail, in a terminal, in a chat box, in a rename field in Finder. On iPhone and iPad it is a system keyboard, so it turns up in the same place every other keyboard does and works in anything that accepts one.
You can transcribe live and watch the text arrive while you speak, or record first and transcribe afterwards, which is what I do on a bad connection. The speech model underneath is ElevenLabs Scribe, across 16 languages.
Around that sits the unglamorous part that decides whether you are still using a dictation app in week three:
- A personal dictionary, so names and jargon stop coming out phonetically.
- Text snippets for the paragraphs you retype constantly.
- 15 voice commands, available in Dutch and English.
- Automatic organisation by topic, so the history is not one long undifferentiated list.
Where does my audio actually go?
Straight from your device to your own ElevenLabs account. Not to a server of mine, not through a relay I operate, not into a queue I could read. You put an ElevenLabs API key into the app once, and from then on the app is a client talking to a service that belongs to you. I am not in the path, which also means I could not hand over your recordings if someone asked me to.
That is the bring-your-own-key model, and it is the whole design, not a setting buried in an advanced tab.
There is a second leak path that dictation apps rarely talk about, and it has nothing to do with transcription: the app phoning home about you. So there is none of it here. No telemetry, no analytics, no crash reporting. Closing the audio path and leaving an analytics SDK running would be a strange kind of privacy.
The rest of it:
- API keys live in the Apple Keychain, not in a preferences file.
- Privacy mode deletes the recording once it has been transcribed.
- Sync runs over your own private iCloud, so your data moves between your devices without passing through an account of mine.
- On iPhone you can put a Face ID or Touch ID lock on the app.
- Rewriting or summarising a transcript with Claude is optional and needs a key of your own as well. Same principle, different provider.
Isn't on-device transcription more private than this?
On one axis, yes, and I would rather say so than argue around it. There are three honest architectures for a dictation app:
- A model that runs on your machine. The audio never leaves. You accept whatever accuracy fits in a download.
- A cloud service run by the app maker. Your audio reaches their servers, and you are trusting a retention policy you cannot inspect. Two parties now hold your speech: the speech vendor and the app maker.
- A cloud model reached through your own account. The audio leaves the device, but only one party receives it, under an account you opened, on terms you agreed to directly, with a key you can revoke this afternoon.
Transcribier is the third. I chose it because I wanted Scribe-level accuracy in 16 languages without becoming a company that stores other people's voices. The honest cost is that the audio does leave your device. If your requirement is that it never leaves the machine at all, take a local-model app with my blessing. That is a good answer to a different question.
What the third option removes is not the network. It is me.
What does it cost, and what does Pro add?
Dictating is free and unlimited. No account, no credit card.
Pro is EUR 4.99 per month or EUR 39.99 per year. It adds:
- Transcribing files you already have, including imported audio and video.
- Reading PDFs aloud.
- Voice isolation, for recordings made somewhere noisier than a desk.
Note what is not behind the paywall: talking instead of typing, every day, in any app. I did not want the core of the thing metered. The transcription usage itself runs through your own ElevenLabs account, so that side is a relationship between you and them rather than something I resell to you with a margin on top. The current details are always on transcribier.com.
Does it behave the same on Mac as on iPhone?
Close, but not identical, because the platforms are not identical. The Mac has a genuine global hotkey and can type into anything on screen. iPhone and iPad get a system keyboard, which is the closest thing iOS has to the same idea, and it reaches every app that shows a keyboard. The biometric lock is an iPhone feature.
What holds them together is your own iCloud. Sync runs over your own private iCloud, and nothing about it involves me. It is one app family rather than three products that happen to share a name, which is the way I build everything: native, on the platform's own frameworks, no Electron.
Why I built it
I dictate a lot, and every option asked me to give up something I did not want to give up. The fast, accurate ones wanted my audio on their infrastructure. The ones that kept my audio at home were limited to whatever model I was willing to download and run. I wanted a different combination: the good model, my account, no middleman.
Building it also settled a question I had about the shape of a small software business. I did not want to run servers that hold recordings, because that is a liability which grows quietly and never stops. Bring-your-own-key removes it at the root. It is a slightly harder sell, since the setup asks one thing of you at the start, and a much easier thing to defend for years afterwards.
You can see the other app I am building on my apps page, and more about how I work on the about page.
Can I get it yet?
Not from the App Store. It is not released, and I would rather say that plainly than let a screenshot imply otherwise. There is a waitlist on transcribier.com, which is where the release will be announced first and where the full feature list lives.
If you are here to see what I make rather than to install something today, Best-I is the other app in progress. Different problem, same rules about your data.