Your audio and your text never leave this machine. Not while you dictate, not afterwards, not ever.
Dictate into email, documents, chat or code, and the words land where your cursor already is. The model runs on your machine, not in someone's data center.
Free to use, with no time limit and no account. Nothing to cancel.
Requires Apple Silicon (M1 to M5) and macOS 15 or later. A connection is needed once, to download the AI model you pick.
Every dictation app says it respects your data. Here are three ways to test whether this one means it.
Once the model you picked has downloaded, transcription never needs a connection again. Pull the plug mid-sentence and the words still arrive.
Point a firewall at the app and you will see everything it ever asks the network for: the one-time model download, the App Store confirming your purchase, and the license check at launch, which is usually answered from the local cache. What you dictate is not on that list. The audio and the text stay on your Mac, always.
Nothing to sign up for and nothing to log in to. No analytics, no usage tracking, no crash reporting of your own. No exceptions.
Mail, Pages, a code editor, a chat window. The text lands in whatever has focus, as though you had typed it there yourself.
Punctuation, capitals and the obvious slips are fixed while you talk, so you are not editing the transcript afterwards.
Pro feature
Say the first word and recording begins. Stop talking and it ends. Nothing to hold down, nothing to click, and it hears when you are done.
Pro feature
Speech is auto-detected across more than 90 languages, and the app's own interface is available in 18.
Drop in an interview, a meeting recording or a video and get the text back, on the same machine, with no upload step. Several files in one go, worked through one after another.
Turn it on, pick a folder once, and every transcription is saved there as its own .txt, named after the recording it came from. Off until you ask for it, and dictation is never written to disk.
Local models including NVIDIA Parakeet and Whisper Turbo, compiled for Apple Silicon rather than streamed from a data center.
Legal drafts, clinical notes, unpublished reporting. Work that cannot go to a third-party server is the reason this app exists.
Written to be useful rather than flattering, including the cases where the other app is the better buy.
Pro adds two things: recording that starts and stops when you speak, and grammar and punctuation fixed by a small model running on your Mac. Everything else is free and stays free.
Manual dictation with any local model, transcription of audio and video files, history and a basic spell check. No account, and this is not a trial.
Hands free recording and automatic cleanup, for a single project or to try properly before committing.
Subscribe on the App StoreThe same two features all year, at $3.33 a month, billed once.
Subscribe on the App StorePay once and keep hands free recording and automatic cleanup, including the updates that follow.
Buy Lifetime on the App StoreThe free tier above has no trial and does not expire. Pro Monthly and Pro Yearly include a separate one-month free trial; Pro Lifetime has no trial, because the App Store does not offer trials on one-time purchases. Everything is bought through the Mac App Store, so canceling and refunds go through Apple.
VocoSafe AI is written and supported by Oleksiy Krotov, an independent developer. There is no team, no help desk and no ticket queue: a support request goes to the person who writes the code.
On the Mac, the app is distributed through the Mac App Store. Apple reviews every update for security before it reaches you, and payment is handled securely by Apple. Every version that has shipped is listed in the changelog, with the date it went out.
