Privacy policy

Effective 24 August 2026 · beta

The short version

Semoru exists to type what you say. Your audio is never stored by us — not on your disk by default, not on our servers, not with any third party. Your transcripts live on your computer. We keep the minimum an account needs: your email and usage counts.

What happens to your voice

On Apple silicon Macs with on-device mode (the default once the model is downloaded), audio is transcribed on your computer and never transmitted. Without on-device mode, audio is sent over TLS to our server in your chosen region (Asia, EU, or US), transcribed in memory by models we run ourselves, and the audio buffer is freed immediately after the text is returned. We do not write audio to disk, we do not log its contents, and we do not use third-party AI providers.

What we store

Your transcripts, history, dictionary, and snippets are stored only in a local database on your computer. Deleting the app deletes them.

What we send for cleanup

The cleanup step (removing fillers, punctuation) processes the transcribed text on our server. If you enable the optional context or style-matching features, up to 200 characters near your cursor or a couple of your recent dictations accompany the request, are used for that single request, and are not stored. Both features have switches in Settings → Privacy.

What we never do

Subprocessors

Your rights

Delete your account from Settings → Account → Delete: every session is revoked and your personal data is scrubbed from our database. Export or ask questions any time: hello@semoru.com. If you are in the EU/EEA, you have the rights described in the GDPR, including access, rectification, and erasure.

Changes

If this policy changes materially, the app will tell you before the change applies to you.