Dictation for macOS · Apple Silicon · $29 once
The Mac dictation app
that never guesses.
It listens only for the languages you picked. It sends nothing anywhere. You can check both in ten minutes.
$29 once · no subscription · no account · no telemetry
Hola Marta, te envío la versión corregida del contrato.
Spanish, because Spanish is on your list.
Or hold Space anywhere on this page. In the app it is the same gesture, in any application on your Mac.
You are saying, in Spanish: «Hola Marta, te envío la versión corregida del contrato»
No microphone was asked for. Nothing was sent. That is also true of the app — the difference is that there it runs on your voice.
The switch no other app has
Recognition is limited to the languages you choose — never guessed from a hundred.
Vocula asks which languages are yours and takes the rest off the table. Not one pinned language you re-pick from a menu every time you switch. Your set, in your order, all at once — the rest are not candidates, so they cannot win.
Same words. Same microphone. One setting.
Hola Marta, te envío la versión corregida del contrato.
Spanish, because Spanish is on your listHola Marta, te envío the corrected version del contrato.
You spoke Spanish. Nothing warned you.Reconstructed for this page, not recorded — the failure is not invented. These models are English-first: let one guess and English is what it reaches for. Of nine Mac dictation apps measured, none offers the setting on the left. They pin one language, or let the model guess from all of them.
“Útil, pero siempre le hablo español y muuuuchas veces el texto lo transcribe mitad en español y mitad en inglés mezclando palabras.”
“Useful, but I always speak Spanish to it and very often it transcribes the text half in Spanish and half in English, mixing words up.”
Mac App Store, Spain — two stars, “Se lía con los idiomas”, April 2026
Three on, the rest off. Change the set with one shortcut — this is Whisper's own list, read from the engine rather than written down here.
Which engine this describes. The switch above is how Vocula's default engine, Whisper, decides. On the second engine, Parakeet, automatic detection works the language out inside the same pass that transcribes it, so your list does not constrain it in the same way. We would rather say that here than let you find it out afterwards. Both engines, and how to pick →
The other half of the pitch
You already talk faster than you type.
That was never the hard part. The hard part is the loop — and a loop with a window in it, or a file to export, or a paste to remember, is a loop you stop using by Thursday. Vocula is the loop and nothing else.
- 01Hold one key, in whatever app you are already in.
- 02Speak — on your own Mac, from a model in your own folder.
- 03Release. The text is at the cursor. Mail, a terminal, a text field.
Two engines. Pick the trade-off yourself.
Whisper Large v3 Turbo default
About a hundred languages, on your Mac's GPU. The engine the language switch above belongs to, and the one to stay on if you mix languages inside a sentence. 1.62 GB.
Parakeet TDT v3
25 European languages, on the Neural Engine. A third of the download and a fraction of the memory — and it cannot transcribe Japanese, Chinese, Arabic or Hindi at all. 468 MB.
Both are local, both are downloaded once, and only the one you select is downloaded. Nothing about the privacy claim changes with either. The full comparison of the two →
“I end up giving much more context and details if I'm speaking out loud compared to typing. Sometimes I might feel a little bit lazy to type one or two extra sentences. But while speaking, I don't really have that kind of friction.”
Hacker News, 7 April 2026
The one number we have measured. On a single Mac — an M5 Pro — across five recorded fixtures of 1.4 to 2.5 seconds, Parakeet returned the text in 57–76 ms and Whisper in 217–230 ms. Both transcribed all five identically. That is short phrases, on that machine, in those conditions; it is not a general speed claim and it is not an accuracy comparison. Your Mac is not that Mac.
There is still no words-per-minute figure on this page and will not be one until we have measured that too, naming the Mac and the engine. Every rival quotes a multiple. None of them says against what.
Not a promise. A command.
Vocula makes zero network connections while you dictate.
Whichever engine you picked, it runs on your own Mac from a file in your own folder — Whisper on the GPU, Parakeet on the Neural Engine. There is no account, no licence server and no crash reporter, and the source is GPL-3.0 and published. Every app here says “private”. This one hands you the command that proves it.
The ten-minute check →$ nettop -p Vocula
bytes_in bytes_out state
0 0 —
$ # still zero. that is the entire product.- No account
- No email address
- No licence server
- No crash reporter
- No telemetry
There is nothing to leak, because there is nothing collected — and that is why the command above returns what it returns.
And this page loads nothing either. No analytics, no tag manager, no fonts from someone else's domain, no embeds — one HTML file and its own stylesheet. Open the network tab, reload, and count. Every other app in the table below sells “private” from a page that reports your visit to Google.
The arithmetic
Vocula is $29, once. Wispr Flow is $15, every month.
Two months of the subscription you are on. Then it stops.
| Vocula | Wispr Flow | |
|---|---|---|
| One person, one year | $29 | $180 |
| One person, three years | $29 | $540 |
| Twenty people, three years | $420 | $10,800 |
Wispr Flow at its published $15 monthly rate, read on 21 August 2026. Billed annually it is $12 a month, so the honest version of the middle row is $432 rather than $540 — still fifteen times this page's price, and still due again next year. Vocula's team figure is a volume licence, paid once — the full pricing.
How Vocula compares to Wispr Flow, superwhisper, MacWhisper and VoiceInk.
| Vocula | Wispr Flow | superwhisper | MacWhisper | VoiceInk | |
|---|---|---|---|---|---|
| Detection limited to the languages you chose | ✓ | — | — | — | — |
| Audio never leaves the Mac | ✓ | — | cloud optional | cloud optional | cloud optional |
| History encrypted at rest | ✓ | — | — | — | — |
| What it costs | $29 once | $15 / month | $249.99 once | €65 once | $29 once |
Read on 21 August 2026 in each app's own source or terms, at their cheapest published rate. The first row describes Vocula's default engine, Whisper — on Parakeet the language set behaves differently, and saying so is cheaper than being caught. If a row has gone stale, tell us and it gets corrected. The full comparison, including every row where Vocula loses →
What Vocula does not do.
You would find all of this out in the first ten minutes. Better that you know before you pay. Unlike Apple's built-in Dictation, none of it involves a server.
- No AI rewriting
- No live preview while you speak
- No custom vocabulary
- macOS 26 or later
- Apple Silicon only
- 1.62 GB first download, or 468 MB
And one that belongs to the faster engine. Parakeet holds to one alphabet per phrase, so an English term inside a Cyrillic or Greek sentence can come back transliterated or dropped, where Whisper keeps it — and pinning a language makes that more likely, not less. It is how the model is built rather than a setting we have not added, and it is sourced rather than assumed. If your dictation mixes languages inside a sentence, stay on Whisper: it is the default, so that costs you nothing. The full mechanism →
Common questions about Vocula
Does Vocula send my voice to a server?
No. Vocula makes no network connections after the model download, which you can verify yourself with nettop -p Vocula while you dictate.
Is Vocula a subscription?
No. Vocula is a one-time $29 purchase with no account and no recurring charge.
Does Vocula run on Apple Silicon?
Yes. Vocula requires an Apple Silicon Mac running macOS 26.0 or later. It does not run on Intel.
Can Vocula type into any application on my Mac?
Yes. Vocula inserts the text at the cursor in any macOS application — Mail, a browser, a terminal, an editor — not only in its own window.
Does Vocula ship more than one recognition engine?
Yes, two, and you pick which one runs. Whisper Large v3 Turbo is the default; Parakeet TDT v3 is a smaller, faster alternative that knows 25 European languages. Both run entirely on your Mac.
How much disk space does Vocula need?
The Whisper engine is a 1.62 GB download and Parakeet is 468 MB, plus 885 KB for the voice-activity detector either way. Only the engine you select is downloaded.
Your voice stays.
The text arrives.
Download for macOS
$29 once · no subscription · no account · no telemetry