Voxlog Privacy Policy

Last updated: 2026-10-08. Applies to the Voxlog iOS app (co.superduperai.voxlog).

App Privacy label: Data Not Collected. Voxlog has no account, no server, no analytics SDK and no network code of its own. The only network use is a download you start:

Speech models (Settings → On-device models → Download): Parakeet (FluidAudio) and the speaker-diarization models, fetched once from Hugging Face; GigaAM (Russian), fetched once from GitHub. The app keeps FluidAudio’s offlineMode on at all times except during that download.
Apple’s speech / language models: installed by iOS itself (Dictation, Live captions, Apple Intelligence) through Apple’s asset system.

Nothing you record, transcribe or summarize is ever uploaded. Turn on Airplane Mode: every feature keeps working.

What lives on the phone

All files are under the app’s Documents/recordings/<YYYY-MM-DD>/ folder (visible in Files and Finder), deleted automatically after the retention window unless starred:

File	Contents
`HH-mm-ss.m4a`	The audio (AAC, mono, 48 kHz).
`HH-mm-ss.txt`	Plain-text transcript, one line per phrase: `[m:ss] S2: text`.
`HH-mm-ss.json`	The same transcript with start times and speaker numbers (for tap-to-seek).
`HH-mm-ss.summary.json`	Headline, key points, to-dos, people — made by Apple Intelligence on the phone.
`day.summary.json`	The same, for the whole day.
`HH-mm-ss.keep`	Marker: exempt from auto-delete.

“Export day” zips exactly these files for AirDrop or Files; nothing else is included.

Third-party code

FluidAudio (Apache-2.0): on-device speech recognition and speaker diarization. It contains no telemetry.

Permissions

Microphone — to record.
Speech Recognition — Apple’s on-device recognizer (iOS 17–25) and live captions.
Reminders — only when you tap “Add to Reminders” on a summary’s to-dos.

Recording other people

Recording third parties requires their consent in most jurisdictions. Voxlog shows the system microphone indicator while recording and cannot hide it.

Live captions

Live captions use Apple’s on-device speech recognizer (or the iOS 26 SpeechAnalyzer). They are shown on screen only, never stored unless you tap “Copy all”, and can be switched off on the main screen. Optional translation of captions uses Apple’s Translation framework on the phone (iOS 18); the language pair is downloaded once by iOS, the text stays local.

Data we collect

None. The App Store privacy label is “Data Not Collected”. There is no crash SDK and no analytics; TestFlight crash reports, if you opted in to share them with Apple, reach us through Apple only.

Contact

info@superduperai.co