
Brief your coding agent like a colleague, not a search box.
Keep the idempotency check as it is, and do not change the existing tests.
When a contract, a client or your employer rules out cloud AI tools, cloud dictation goes with them: your voice, and everything you say, would go to a server you do not control. Dictate Safe runs the speech model on your own computer. On the default setup, nothing you say is sent anywhere.
25 free dictations · one-time purchase · macOS and Windows
An attorney, a nurse practitioner and an engineer, in three cities. Each one says their work out loud, and each room shows one thing that never leaves the computer.

Deposition outline
Draft for Wednesday, 9:30 AM
C. The revised contract
Question 10. Who drafted the first version of the agreement?
Question 11. Who else reviewed the draft before it was signed?
Question 12. When did you first see the revised contract?
Tomorrow’s questions, worked out loud the night before, in the words she will actually use. The speech model runs on her computer, so what she says about the case goes nowhere else: no account, no telemetry, nothing sent.
Local is the default: the speech model and the filler-word removal both run on the computer, and the app always shows which mode it is in.

A minute between rooms is enough to say a note, not to type one. Every dictation is kept only on her computer, and when she deletes one, its text and its recording are erased there, not just hidden.
Delete one dictation or many: the text and the recording are erased on the computer, and the space they used is overwritten.

The client’s rules say no recordings and no cloud tools, so the notes come from him, straight after the call, in his own words, into a note in Dictate Safe’s own Scratchpad. There is no meeting recorder: it listens only while he dictates. With Offline Lock on, it makes no connection at all, and the switch shows on every screen of the app.
Offline Lock refuses model downloads, update checks, signing in and cloud providers while it is on, and says so in plain words.
The default sends nothing. The other two exist for people who want cloud speed, only with a provider and a key they chose themselves.
Nothing leaves your computer. The speech model runs on it, and the filler words are removed on it.
Only the text goes to the cloud provider you picked, for AI cleanup. Your voice stays on your computer.
Your recording goes to Groq, with your own key. We never see the audio, the text or the key.
Whether a tool is allowed is your employer’s or your client’s decision. These are the facts to make it with, each one something they can check.
More for a security review on the Security page and in the Privacy Policy.
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

Keep the idempotency check as it is, and do not change the existing tests.

