Privacy
MaxTalk has no servers of its own and collects no usage data. Where your words go depends on the speech recognizer and AI provider you choose.
Your voice
- Local Whisper (default): recordings are recognized on your Mac and sent nowhere.
- Cloud recognition (optional): recordings go to the speech provider you chose in AI Models.
- Recordings are not kept by default; only if you turn on “Keep recordings”, or a recognition fails and needs a retry, are they stored on your Mac.
What goes to your AI provider
Only when AI is used (the AI key, or AI cleanup for dictation) does a request go from your Mac straight to the provider you chose, never through MaxTalk. It contains:
- the recognized text, plus the current time and time zone;
- the name of the app in front, and whether the cursor seems to be in a text field;
- the text you selected (turn off “Read selected text” in Settings › Privacy);
- your vocabulary, and the names of your Shortcuts (turn off in Settings › Actions; never what they do).
Your provider's privacy policy applies to it. The decision model (Jev) in Labs is off by default; when on, it sends similar content to TypeSafe (or the gateway you chose, such as OpenRouter).
Kept on your Mac
- API keys: only in the macOS Keychain; never written to files, logs or exports.
- History: on your Mac; set how long it is kept, export it, or clear it at once in Settings.
- Usage: the model, tokens and time of each call, without content, on your Mac only.
- Logs: for troubleshooting, redacted, without keys or what you said.
Other network requests
- Downloading the speech model connects to Hugging Face (or the hf-mirror.com mirror when that fails).
- When you press a talk key, MaxTalk opens a connection to your AI provider ahead of time (no key, no content) so the request is faster.
This website
This site is hosted on Cloudflare Pages. It uses no cookies or analytics; it only remembers your language choice in your browser.