Nothing here was built because a competitor had it.
Dictate Safe does one job: dictation you can use anywhere, without your voice leaving your machine. Everything on these pages either makes that work, or makes it survive the way you actually work all day. The things we chose not to build are here too.
Hold one key. Talk like a person.
It’s already typed.
Not a window you open, not a box you paste out of. You hold a key in whatever app you were already in, say the messy version out loud, and finished text lands where your cursor was. The words start appearing while you are still talking.
Windows build is signed and notarised.
Docs are through review.
The same shortcut works in a notes app, a terminal and a chat app like Claude: one cursor, a different sentence in each, no switching tools.
The voice note you keep meaning to listen to.
Someone sends nine minutes of talking on WhatsApp. A recording from a call sits on your phone. Right now those are a chore you postpone, because listening is slower than reading and you cannot search a waveform.
Drop the file in and it becomes text you can skim, search and quote. It runs through the same on-device engine as live dictation, so a recording of a private conversation does not have to be handed to a transcription service to become useful.
A WhatsApp voice note dragged from Downloads onto the Import page: transcribed on this Mac, and saved to History as well.
One pass is free and on by default.
AI cleanup is there if you want more.
Most people never touch this. The free pass alone produces the everyday quality on this page. Turn on AI cleanup if you change your mind mid-sentence, or if names from your Vocabulary keep coming out slightly wrong.
The free pass
Plain code, no AI involved, nothing sent anywhere. It runs instantly, works in airplane mode, and is what produces the everyday quality you will see.
- Strips true filler sounds: um, uh, er, ah, erm
- Spoken punctuation, so “full stop” and “new paragraph” become real formatting
- Capitalisation where it belongs
This is the path in use for the speed figures on this page. No model, no key, no account.
AI cleanup, on-device
Turn it on and an AI model on your own machine does the two jobs a plain transcript cannot: it follows you when you change your mind mid-sentence, and it puts names and terms back the way your Vocabulary spells them. It never rewords what you said.
- “Meet on Monday, no wait, Wednesday” arrives as Wednesday
- Free, no account, no API key
- Only runs when there is something to fix, then takes a few seconds on an older machine
- Needs a machine with 16 GB of memory
Off by default. On or off, your voice never leaves your device either way.
AI cleanup, your cloud key
The same two jobs under the same rules, done by a cloud model through your own key instead. It is a speed option, for an older machine or one without the memory for the on-device model.
- Works on any machine
- Faster on an older one
- Runs through your own Groq or OpenAI key, at cost, billed to you by them
Off unless you switch it on. This one is a cloud feature: the text goes to the provider you chose. We never see the audio, the text, or the key.
Switching on AI cleanup in Settings, then dictating into Claude: the change of mind in the middle of the sentence is resolved on this Mac before the text lands.
The same icon holds your tasks and an assistant.
The floating icon that starts a dictation also shows your reminders, takes a new task when you say one, and opens a small chat with the Assistant. Both work on your computer by default. Each has its own page, with recordings of the real app.
Tasks and reminders on the floating icon
Add a task by voice, reminders that do not pile up, goals, a board, a calendar and a mind map.
See tasks AssistantBetaAn assistant that works on your computer
Ask for today's tasks, plan the week, draft an email, and approve every change with one press.
See the AssistantBuilt for people who talk to an AI all day
The prompt library, Code Mode, keeping the computer awake through a long job, and the everyday parts you notice on day thirty.
See power use OfflineHow the offline engine works
Where your voice goes, one dictation with the Wi-Fi off, everything that ever goes online, and Offline Lock.
See how it stays offlineWhat we decided not to build.
A features page that only lists what a product does is half a description. These are real capabilities that competitors ship, and here is why each one is missing.
Meeting recording
Recording other people without telling them is illegal in twelve US states alone, and tools that do it quietly are exactly what gets software banned by the IT teams this product is built for. Import an audio file you already have and are entitled to, and you get the same text without the liability.
Reading your screen
Some dictation tools capture what is on your display to sharpen their guesses. It works, and it is also a second stream of your private information leaving the machine. We took the accuracy loss instead.
Every language
Dictate Safe is English only today. The models it runs are English-specific builds, which is part of why it is accurate at small sizes that fit on a laptop. If you need to dictate in another language, this is the wrong tool right now and we would rather say so here than after you have paid.
The specifics, in one place.
Try the part that matters first.
Twenty five dictations, no card and no account. It is the same app you would buy, running the same way it will on day thirty.
One time · 2 devices · 30 day refund
