Speak.It’s written.

No. Speech-to-text runs on your device using local models, for dictation and for meetings. Ambient does not upload or save your recordings. When an AI feature needs a model, only text is sent, never audio.

With Pro, yes. The dictation key decides from what you say whether to write (polished text pasted where your cursor is), answer (a short answer with sources), act (calendar events, reminders, opening apps, running your Shortcuts) or recall something you said or saw. It also takes meeting notes and, if you switch it on, offers one next step at the right moment.

Only if you let it. Everything that looks at your screen is off until you turn it on, one capability at a time. It reads text, not pixels, keeps it on your Mac encrypted (screen memory for 7 days by default), and skips password managers, banking, private windows and messaging apps out of the box. One key pauses it, and the HUD is invisible in screen shares.

Only the text a request needs, at the moment it needs it: the words to polish, your question, or the snippet you asked about. Secrets such as card numbers, keys, passwords and codes are redacted on the Mac first. The network ledger in the app shows every request and exactly what was sent. On-device mode uses Apple Intelligence with no web search, so nothing leaves at all.

No. Ambient transcribes the call on your Mac from your microphone and the Mac’s audio, so it works with Zoom, Meet, Teams, FaceTime and Slack huddles with nothing to invite or connect. The audio is never saved. You get rolling notes, live nudges, and afterwards a summary, decisions, action items and a follow-up email draft.

No. Ambient finds your own things where they already are: open windows, a two-minute in-memory buffer, encrypted screen memory, Spotlight (files, Mail, notes), Calendar and your dictation history. Drive and Gmail are searched in the browser you’re already signed in to.

Dictation does. Once a model is downloaded, it works with no internet connection: on a plane, on a train, or behind a strict firewall. Answers that search the web need a connection.

Dictation runs on macOS, Windows and Linux. The agent features (answers, actions, meetings, suggestions and recall) are on macOS for now.

Unlimited dictation, forever. It runs on your computer, so it costs us nothing to give away. Every install also gets 7 days of everything. After that, Free includes 25 AI uses a week and 2 meetings a month. Pro removes both limits.

One polished dictation, one answer, one action, one Ask or one meeting summary each count as one. The 25 a week are shared across all of them and reset every Monday (UTC). Plain dictation never counts.

Yes. Every new install gets 7 days of everything, with no card and no account. After that you stay on Free, with unlimited dictation and a weekly allowance of AI uses.

Coming soon. Ambient will operate apps on screen the way a person does, so “make a Linear ticket for the login bug” works without connecting Linear. It will ask before anything that sends, deletes or buys.

Any app where you can type. Ambient pastes the text wherever your cursor is: email, chat, docs, code editors, the terminal.

After checkout you’ll get a license key. Open Ambient → Account, paste the key and click Activate. The key is verified on your device and refreshes automatically while your subscription is active.

Yes. Manage or cancel your subscription from Ambient → Account → Manage billing. Pro stays active until the end of the period you paid for.

Ambient is built on Handy, an open-source dictation app by CJ Pais, released under the MIT license. The transcription engine is open source; Ambient adds the agent, licensing and support on top.