Nexus AI LogoTalk To Me

Privacy Policy

Talk To Me — Audio to Text · Effective and last updated: 21 August 2026

This policy explains how the Talk To Me Android app (listed on Google Play as “Talk To Me: Audio to Text”) handles your information. Talk To Me is developed by Nexus AI Technology Labs, a brand of AI Nexus Consulting FZ-LLC. For any question about this policy or your data, contact us at saksham@nexusai.world.

The short version

Talk To Me does not collect anything. There is no account, no analytics, no advertising, no crash reporting and no tracking of any kind. Your recordings and transcripts stay on your phone.

Data safety at a glance

  • We collect nothing. There is no account, no backend, and no path by which your recordings or transcripts reach us.
  • We share nothing. We operate no analytics service and no advertising, and we hand your data to no one.
  • Nothing in the app reports to anyone. Talk To Me bundles no analytics, crash-reporting or diagnostics component — no Firebase, no Google Play Services, no ML Kit. Its only outbound request in its entire lifetime is the one-time speech-model download described below.
  • All transcription is on-device. Speech becomes text on your phone. Your audio is never uploaded, to us or to anyone else.
  • You are the only export route. The single way a transcript leaves the device is you copying it or sharing it yourself.

What stays on your phone

Everything you create with the app:

  • Recordings. Audio you record, and audio you share or open from another app, is copied into the app’s private storage. Other apps cannot read it.
  • Transcripts. Speech is turned into text entirely on your device. Transcripts are stored in a database that is encrypted at rest, under a key held in the Android Keystore.
  • Your settings. Kept on the device and excluded from cloud backup.

None of this is uploaded, and none of it is visible to the developer.

The one time the app uses the network

The first time you transcribe something, the app downloads its speech model — about 660 MB, once — over HTTPS from the sherpa-onnx author’s public model repository on Hugging Face. The request contains nothing about you: it is an ordinary file download, and every file’s integrity is checked with a SHA-256 hash before it is used.

That is the app’s only network request. Plaintext traffic is disabled outright. After the download, transcription works with no connection at all. You can hold the download until you are on Wi-Fi from Settings, and you can remove the model again at any time.

Permissions

Talk To Me declares seven permissions:

  • RECORD_AUDIO — requested when you start a recording, and used only while a recording is running.
  • INTERNET — the one-time speech-model download above. Nothing else in the app makes a network request.
  • POST_NOTIFICATIONS — asked for once a recording is under way, so the app can tell you when a transcript is ready. Declining costs you nothing; the transcript still appears in the app.
  • FOREGROUND_SERVICE with FOREGROUND_SERVICE_MICROPHONE — so a recording survives you leaving the app. It is held only while a capture is in progress and stops the moment the recording ends.
  • FOREGROUND_SERVICE_DATA_SYNC — so a long transcription or the model download keeps running when backgrounded.
  • ACCESS_NETWORK_STATE — to ask whether the current network is metered, which is what makes the “download over Wi-Fi only” setting work. It reads the connection type, not your traffic.

Two more are merged into the installed app by Google’s androidx.work library, not declared by Talk To Me: WAKE_LOCK and RECEIVE_BOOT_COMPLETED, which that library uses to keep scheduled work running. Android also generates a signature-level permission named after the app itself (ending DYNAMIC_RECEIVER_NOT_EXPORTED_PERMISSION), which exists so the app’s own internal broadcasts are not exposed to other apps. If you inspect Talk To Me’s permissions in Android settings, this is why those three appear there.

The app asks for no storage permission, no location, and no contacts. Audio shared into it arrives as a content URI, and files it shares out go through a temporary provider you invoke yourself.

When something does leave your phone

Only when you send it. Copying a transcript, or sharing it as a .txt file, hands it to whichever app you pick. What happens to it then is governed by that app’s policy, not this one.

Backups and device transfer

Talk To Me opts out of Android’s automatic backup entirely, and its recordings, transcripts, downloaded model and settings are additionally excluded from both Google cloud backup and direct device-to-device transfer. Nothing you create in the app is copied to Google’s servers, and nothing follows you to a new phone.

The transfer exclusion is deliberate rather than an oversight. The transcript database is encrypted under a key that is bound to the hardware of the phone that created it, so the key cannot move even though the file could. Carrying the database across on its own would hand the new device a history nothing could open.

Deleting your data

  • Delete any single transcript from its own screen or from History. Its recording goes with it.
  • Settings → About → Delete everything removes every transcript and every recording at once.
  • Turn off Keep recordings in Settings and each recording is deleted as soon as its transcript is written.
  • Uninstalling the app removes all of it, including the downloaded model.

Children

The app is not directed at children and collects no personal data from anyone.

Third-party components

We do not sell, rent, or share your information with any third party. There is no advertising in Talk To Me, and no analytics of any kind — we have no console, no dashboard, and no way to see anything about you or your use of the app.

For completeness, the notable third-party components inside the app are: sherpa-onnx and ONNX Runtime, which run the NVIDIA Parakeet speech model on-device; SQLCipher, which encrypts the transcript database; and Google’s AndroidX/Jetpack libraries (including androidx.work for background transcription). None of them is an analytics, advertising or crash-reporting component, and no Google Play Services, Firebase or ML Kit component ships in the app at all. The full list, with licences, is in the app under Settings → About → Open-source licences.

Changes to this policy

If this policy changes, we will post the updated version on this page and revise the effective date above, and the updated text ships in the app itself. Material changes will be highlighted here.

Contact

Nexus AI Technology Labs (a brand of AI Nexus Consulting FZ-LLC). Questions about this policy or your data: saksham@nexusai.world.

← Back to Talk To Me overview