Voxlog: privacy policy
Last updated November 2, 2026. Checked against the app's code on that date.
The short version
Audio, transcripts, captions and summaries are files in the app's own folder on your iPhone and are never sent to us. The app has no analytics and no server of ours; it only downloads speech models when you tap Download.
- Account: none. The app has no sign-in and no account.
- What reaches us: nothing. We run no server this app reports to.
- Selling or sharing data: never. No advertising, no ad identifiers, no data brokers.
What stays on your device
- Audio (Documents/recordings/<date>/HH-mm-ss.m4a, AAC mono 48 kHz)
- Transcripts (.txt and .json with timestamps and speaker numbers)
- Live-caption text files (captions-HH-mm-ss.txt), saved when a recording stops
- Summaries (.summary.json, day.summary.json) and .keep star markers
- Settings in UserDefaults (engine, language, retention days)
- Downloaded models in Application Support (GigaAM excluded from backup)
Other services the app connects to
Each of these receives only what is listed, plus the IP address any network request carries. Services run by others handle it under their own privacy policies.
| Service | What it receives | When |
|---|---|---|
huggingface.co | Standard HTTPS file requests for Parakeet ASR (about 220 to 650 MB depending on variant) and speaker-diarization models (about 30 MB); no user data | Only after tapping Download in Settings > On-device models; FluidAudio offlineMode is on at every other time |
github.com (release fortunto2/gigaam-v3-coreml v3-rnnt-int8) | HTTPS downloads of GigaAM Russian model files, about 230 MB; no user data | Only after tapping Download with GigaAM selected |
Apple (iOS asset system / Siri speech servers) | iOS downloads dictation, Translation and Apple Intelligence language assets itself. Edge case: if no on-device dictation model exists for any candidate language, the code falls back to requiresOnDeviceRecognition = false, which lets Apple's server recognizer receive the audio | Asset downloads when iOS decides; server fallback only when the device has no on-device dictation model for any candidate language |
Permissions
iOS asks you before the app can use any of these, and you can change your answer in Settings at any time. The app keeps working without them, minus the feature that needs it.
- Microphone: Recording
- Speech Recognition: Apple's recognizer for transcripts and live captions
- Reminders (full access): Only asked when tapping Add to Reminders on a summary
- Background audio: Keep recording with the screen off
- Live Activities: Dynamic Island / Lock Screen recording indicator
Analytics and crash reports
None. The app contains no analytics code of any kind.
No crash reporter in the app. If you chose to share analytics with app developers in iOS Settings, Apple sends us anonymous crash logs; you can turn that off under Settings > Privacy & Security > Analytics & Improvements.
Speech recognition fallback
Transcription runs on the device. One edge case: if iOS has no on-device dictation model for any language the app tries, the app lets Apple's speech recognizer handle it, and Apple's servers then receive that audio under Apple's own privacy policy. Installing the dictation language in iOS Settings, or using the Parakeet or GigaAM models inside Voxlog, keeps it on the device.
Children
The app is for a general audience. It asks no one, children included, for a name, email, age or any other personal detail.
Deleting your data
Everything the app keeps lives on your device. Delete the app and iOS deletes its data with it. We hold no account or profile of you, so there is nothing on our side to export or erase; if you think we hold something, write to us and we will check and answer plainly.
Who we are, and how to reach us
Voxlog is made by SuperDuperAI and published on the App Store under the developer account of its founder, Rustam Salavatov. The company behind the brand is SuperDuperAI, Corp., a Delaware public benefit corporation, 1207 Delaware Ave #4241, Wilmington, DE 19806, USA (details on about). Questions about this policy or about your data: [email protected]. Help with the app: Voxlog support.
If the app starts handling data differently, this page changes first and the date above moves. The policy for the website itself is here.