What leaves your Mac

This page describes normal processing and component downloads. The terms for voluntarily sharing data are in the privacy policy.

Processing on your Mac

Recognition, translation and text processing run on your Mac, without sending them to the cloud. Transcripts are kept in local history. You choose whether meeting audio is saved to disk.

Smart Keyboard (a paid feature) sees your key presses so it can fix a word typed in the wrong layout or with a typo — and only with the Accessibility permission. Key presses and typed text are not recorded in history or logs and never leave your Mac. To correct keyboard layouts, Accessibility reads the actual field text before the cursor back to the sentence start or the number of previous words set in settings, up to 16,384 UTF-16 units. If the required start is unavailable, no correction is made. The snapshot is forgotten after the decision; the last correction is kept for undo. Individual letter correction may also read up to 64 characters on each side of the cursor. Search boxes, address bars and password fields are never read. «Is this a word?» is answered by the macOS spell checker and dictionaries built into the app, on your Mac. With «Fix typos using the sentence» on, the possible fixes of a typo are weighed by the local term judge model: it sees only the words of the current sentence and those fixes, in memory. Nothing of this is written to disk or to the log. The list of words you explicitly ask to leave alone is stored only on this Mac.

Network requests

Sign in, sign up, ask for a token — our own server. Sign-up carries your email, your password and a signature made with the install key.

When you are signed in, the sign-in, each licence-token refresh (once a day, or after four hours of use) and the start of a trial also carry your macOS language, the app version and how long the app was in use since its previous refresh; the trial start also carries a one-way fingerprint of the Mac’s hardware (a hash of its hardware ID; the ID itself is not sent). The server records the IP address, country and city the request came from. This keeps it to one trial per Mac and at most five devices per account, and shows when one account is used by several people. IP address and city are kept for 90 days, the rest for as long as the account exists.

Update check — cdn.verbaide.com/downloads/appcast.xml, once a day and when you press the button in About. It is a list of versions: the app only reads it and sends nothing about you. You can turn the automatic check off with a checkbox in About. A new version is installed only when you press Install.

Translation dictionary catalogue — cdn.verbaide.com, when Settings are open or when the language screen of the first-run setup is open.

Translation dictionary files — cdn.verbaide.com, when you download one.

Speech-model registry and models. The recognition/parakeet.json pointer and its manifest on cdn.verbaide.com are checked at launch, at most once a day, and only during the trial, on a paid plan, while first-run setup is open (the trial starts on its first step), or while our model is still installed and the app has not yet confirmed that your plan has ended. Offline, a cached or bundled manifest is used. During the trial and on paid plans the main model is our fine-tuned build, downloaded from the CDN. On the free plan the app uses NVIDIA’s original Parakeet model, downloaded from huggingface.co without asking the registry. When your plan changes, the model you no longer need is removed and the other one is downloaded. The CTC term model (98 MB) downloads from Hugging Face in the background on first dictation. An installed model is updated with your agreement.

Speaker-labelling model. Downloaded from huggingface.co only when you press Download. Speaker labelling is experimental.

Local language-model catalogue and manifest — cdn.verbaide.com/intelligence. Metadata only.

Local language-model weights — huggingface.co/Qwen/…, when you start that download.

Term judge model — cdn.verbaide.com/intelligence/term-judge/… (1.8 GB), only on a paid plan or the trial; it downloads by itself in the background. There are two triggers: the start of a dictation, when dictation improvement is on and your lists hold at least one term; and Smart Keyboard, when it starts listening to the keyboard (the feature is on and Accessibility is granted). The model is not downloaded at app launch, and never on the free tier. Both triggers use the same address.

The dictionary, language-model and term judge requests carry no cookies and no authorization header.

Only with your consent

Usage counters. Numbers only — never text. The list: how many dictations and in which language or language pair; how long you spoke; how many characters came out (the count, not the characters); how long recognition, translation and text processing each took; which paid feature was blocked; and your country, taken from the system region.

Two doors: by session when you are signed in, and when you are not, by a stable device id plus a signature made with the install key. If you did not allow statistics, none of this is sent.

The free tier

Dictation and term dictionaries are free forever, with no word cap and no card. Download the speech models first; afterwards dictation works offline. Component catalogues fetched while Settings are open and the speech-registry check at launch do not depend on your plan.