TalkAfter Privacy Policy
Last updated: 2 November 2026
TalkAfter turns spoken conversation into text on your phone and highlights phrases linked to strengths and concerns. This policy explains what the app handles, where it goes, and the choices you have.
In short: the app has no accounts, no ads and no analytics. Your conversations, transcripts and voice profiles stay on your phone. Nothing is sent to us. Two things can involve another company: the speech model downloads, and Google's speech recogniser if you choose it.
What the app handles
| What | Why | Where it stays |
|---|---|---|
| Microphone audio | To turn speech into text while you are listening | Processed as it arrives; not saved, unless you turn on "Keep audio for playback" |
| Kept audio (optional, off by default) | To play a conversation back with the text following along | Only on your phone, in the app's private storage; deleted with its conversation |
| Recordings you choose to analyse | To turn a saved recording into text | Read on your phone and never uploaded; copied into the app only if "Keep audio for playback" is on |
| Transcripts and highlights | So you can review a conversation later | Only on your phone, in the app's private storage |
| Speaker names you type | To label who said what | Only on your phone |
| Voice profiles (optional) | To recognise the same voice again, and for Self-analysis mode | Only on your phone |
| Your own phrases, dismissed phrases, settings | Your preferences | Only on your phone |
Voice profiles are a numerical summary of how a voice sounds. Some laws treat them as biometric data. The app makes one only when you save a voice or set up your own voice, and you can delete any of them in Settings → Known voices. When a conversation ends, voice data for speakers you didn't save is discarded.
Who else is involved
- Offline engine: speech recognition runs entirely on your phone, using speech models (about 160 MB) that are downloaded once. The English model comes from Google Play, which downloads it right after the app is installed; Google Play's terms and privacy policy apply to that download. The speaker model, the optional Whisper model and the French and Chinese models download from Hugging Face when you ask for them, using Android's download manager. Google Play and Hugging Face see an ordinary download request from your phone, including its IP address. No audio or text is sent.
- Google recogniser (optional): if you choose this engine, Android's speech service, provided by Google, does the recognition. Depending on your phone and the "on-device" setting, audio may be sent to Google. Google's privacy policy applies to that: policies.google.com/privacy
- Sharing: if you tap Share (PDF or text), the app hands the file to the app you pick. What happens next is up to that app and the people you send it to.
We (the developer) never receive your audio, transcripts, voice profiles or any usage data.
The people you record
Conversations include other people. Recording someone without their consent is illegal in many places. Tell everyone before you start. Self-analysis mode keeps only your own voice and discards everyone else's lines without saving them.
Keeping and deleting
- Delete a single line, a whole conversation, or a known voice at any time.
- Settings → Privacy can delete old conversations automatically after 7, 30 or 90 days. It is off unless you turn it on.
- Settings → Privacy → Keep audio for playback is off unless you turn it on. Turning it off offers to delete the audio already kept; deleting a conversation always deletes its audio. Kept audio is not included in backups.
- Uninstalling the app deletes everything it stored. The app opts out of Android cloud backup and phone-to-phone transfer, so nothing is copied off the phone automatically.
- Settings → Back up to a file saves your conversations, voices and settings in one password-protected file, in a place you choose. Where that file goes afterwards is up to you.
- Settings → Privacy → App lock can require your fingerprint or screen lock to open the app, and hides conversations from the recent-apps preview.
Children
The app is not intended for children under 13 and does not knowingly collect information about them.
Not a diagnostic tool
Highlights are rough signals based on words and tone, not a psychological assessment of anyone.
Changes
If this policy changes, the new version will be posted at this address with a new date.
Contact
Questions: support@talkafter.com