Privacy Policy
Overview
Tonota does not collect your recordings, transcripts, or AI tool content on a developer server. Recording, transcription, and AI processing run locally. Your data can move between your paired Apple Watch and iPhone, or leave the app when you choose to export or share it.
What data Tonota accesses
| Data | Purpose | Where it goes |
|---|---|---|
| Microphone audio | Recording voice memos on iPhone or Apple Watch | Stored on the recording device; Watch recordings transfer to the paired iPhone |
| Transcripts, titles, folders, AI tool instructions and results | Creating and organizing memos and running local AI tools | Stored locally on the iPhone |
| Optional exports and shares | Saving transcripts and audio to a chosen folder or using the iOS share sheet | Sent to the destination you choose |
Watch recordings are kept on the Watch until they can transfer to the paired iPhone. Tonota does not upload them to its own server. If you choose an export folder synced by iCloud Drive or another file provider, or choose a sharing destination, that service handles the copy under its own privacy terms.
What Tonota does NOT do
- No user accounts or sign-in
- No analytics or crash reporting to third parties
- No advertising identifiers
- No developer server receiving recordings, transcripts, or AI tool content
- No network connection required for transcription or AI inference after model setup
Tonota connects to model hosts when you download or update a model. After a model download fails, its network check may send a HEAD request to captive.apple.com and huggingface.co to diagnose connectivity. These requests do not include your recordings or transcripts; the hosts can receive ordinary connection information such as your IP address.
Third-party dependencies
WhisperKit (open-source, MIT licence) runs transcription on-device using Apple's Core ML framework. Model weights are downloaded from Hugging Face when you choose a model; transcription does not send audio to the model host.
mlx-swift-lm (open-source, MIT licence) powers Tonota's on-device AI features — translating, summarizing, extracting to-dos, rewriting, polishing, and any custom instructions you create. It runs entirely on your device and does not send your recordings or transcripts to any server during inference.
Apple Foundation Models is an optional on-device language model provided by Apple as part of Apple Intelligence (available on supported devices). When you choose it as your AI model, the same AI features run locally using Apple's built-in system model — your recordings and transcripts are not uploaded by Tonota. As an Apple system feature, its use is also governed by Apple's own privacy terms for Apple Intelligence.
Your rights
Tonota keeps your memos in local app storage until you delete them in the app or remove the app. You can edit transcripts and titles, and export or share content using the app's controls. If you enabled automatic export, Tonota also writes transcript and audio copies to your chosen folder and attempts to remove its matching exports when you delete a memo while that folder remains available. You can inspect or delete those files directly in the chosen folder. Removing Tonota does not remove copies in iCloud Drive, another file provider, or a sharing destination; those services may retain them under their own terms.
Contact
Questions? Email janewu2005@gmail.com