Dictid

Privacy

What leaves your device, and when.

Last updated 19 November 2026

Recordings

Audio is written to the device and stays there. Transcription runs on the device, so no recording is uploaded in order to be transcribed. Audio leaves the device only when you archive it to your own cloud storage, export it, or dispatch a summary, each of which you trigger yourself.

Voices, and how speakers are told apart

Working out who spoke when happens on the device. The model that does it is downloaded once and then runs locally, the same arrangement as transcription, and the voice measurements it works from exist only while a recording is being processed. They are not saved, not sent anywhere, and not used to recognise you or anybody else in another recording. What is kept is the part you can see on the transcript: which speaker spoke, when, what they said, and the label they were given.

Speakers are labelled from what people call each other in the conversation and from who the calendar says was invited. Dictid never asks you to record a voice sample, and never builds a voiceprint to match anyone against later. That is deliberate rather than incidental: a stored voiceprint is a biometric identifier under laws such as the Illinois Biometric Information Privacy Act, and this feature was built so that none is ever created. If that ever changes, this page changes first and the app asks before it collects anything. Speaker separation can be turned off altogether under Settings.

Your model keys

If you add a Claude, Gemini or OpenAI key, it is stored in the device keychain and sent only to that vendor, directly from the device. It is never proxied through us and never stored on our servers. Your transcript is sent to that vendor only when you ask for a summary or use the chat.

Your account

An account exists only to hold a subscription and to scope archived recordings to you. We store your email address, a password hash, and your subscription status. Passwords are hashed with argon2id; we never see the password itself.

Archived audio

Archived recordings are stored under your account and are readable only by you. They are not analysed, indexed or used to train anything. If your subscription lapses you can still read and delete what you have already archived.

What we do not do

No advertising identifiers, no third-party analytics in the apps, no selling or sharing of recordings or transcripts, and no human review of your audio.

Getting your data out, or deleted

Every recording can be exported from the app as audio, text, Markdown or WebVTT. Deleting a recording removes it from the device and, if archived, from storage. Delete account, in your account or under Settings › Account in either app, closes the account and deletes everything archived under it. It takes effect straight away and cannot be undone. Recordings held on your own devices are not touched, because those are files on the device. If an account cannot be deleted that way, write to privacy@dictid.com.