Tetatet: Offline Transcribe

Tetatet records your sessions and turns them into text — entirely on your phone.

No server. No account. No analytics. The audio, the transcript, the client names and your notes never leave the device, because there is nowhere to send them. Confidentiality here is a property of the architecture, not a promise in a policy.

Made for people whose conversations are privileged: therapists, psychologists, lawyers, coaches, researchers.

WHAT IT DOES

• Records a full session in the background — the screen can be off, 50+ minutes is fine
• Transcribes speech to text on the device, with timestamps
• Tells speakers apart automatically; rename them to "Therapist" and "Client"
• Client cards: every session with a person in one place, with audio and transcripts, and how long it has been since the last one
• Imports an existing audio file (m4a, mp3, wav and others)
• Text notes and short voice notes on every session
• Search across the whole archive, within one client, or inside a single transcript
• Exports a transcript (TXT, SRT) or a client's full history — only when you ask it to

PRIVACY, IN DETAIL

• Everything is processed on the device. Turn on airplane mode and transcription works exactly the same
• The app uses the network once: to download the speech models. Nothing about you or your sessions travels with that request
• No advertising, no analytics, no crash-reporting SDK, no third-party services
• App lock with fingerprint, face unlock, or your device passcode
• With the lock on, the app's content is hidden from the recent apps screen and cannot be screenshotted
• Initials mode shows "A.K." instead of full names in lists
• Session data is excluded from Google backup, so it never reaches a cloud you did not choose

BEFORE YOU START

• The speech models are a one-time download of about 600 MB. Use Wi-Fi; afterwards they stay on the device and no download is needed again
• Because nothing is backed up, nothing can be restored. If you lose or reset the phone, or uninstall the app, the recordings are gone. Export whatever you need to keep
• Automatic transcription is not perfect, and speaker labels can be wrong. Check the text against the audio before you rely on it as a record
• Requires Android 12 or newer on a 64-bit device. An hour-long session takes several minutes to transcribe, depending on the phone

LANGUAGES

Interface and speech recognition in English, Chinese, Hindi, Spanish, French, Arabic, Bengali, Portuguese and Russian.

RECORDING OTHER PEOPLE

Consent rules differ by country and by profession, and are often stricter than people expect. Obtaining consent and meeting your professional obligations is your responsibility.

Privacy policy: https://github.com/pushkadev/Tetatet/blob/main/PRIVACY.md