Soumari Privacy Policy
Development privacy draft, reviewed against project sources on 16 October 2026. Final release behavior and disclosures still need verification.
Summary
Soumari is a local-first sound recorder with local transcription and AI summaries. Recordings stay on the device unless the user chooses to export or share them. These features do not upload recordings to a cloud AI service.
Data Collected
The current project documentation describes no account, analytics or recording-upload service. This is an implementation description, not a completed release audit.
Data Stored Locally
Soumari stores recordings, recording metadata, markers, notes, transcripts, AI summaries, downloaded models and settings locally. Temporary playback or processing files may be created in app cache.
Optional Model Downloads
Whisper transcription and Gemma summaries require local model files. Optional downloads use network access and contact their model-hosting providers, which receive ordinary connection information such as an IP address. The app's network-access setting is off by default. Model downloads do not require uploading recordings. Gemini Nano availability and model provisioning depend on the device and Android services. Applicable model terms may need to be accepted before downloading.
Data Shared
If the user exports or shares audio, a transcript, summary or other generated document, the selected destination receives that content. Its handling is governed separately by that destination.
Permissions
Microphone access is required to record audio. Notification or storage-related permissions may be used only if the final Android implementation needs them.
Children
Soumari is not planned as a child-directed app.
Deletion and Export
Users should be able to delete recordings and intentionally export or share recordings from inside the app.
Contact
For privacy questions, email jamps3@gmail.com.