VoxVault is designed to be private by default. This policy explains what the app and this website do and do not do with your data.
Summary
What data VoxVault handles
When you use VoxVault, the following are created and stored on your iPhone or iPad (and, like any app's data, included in your encrypted device backup unless excluded in Settings):
- Audio recordings you choose to capture or media you choose to import
- Transcripts generated on-device
- Summaries and action items generated on-device
- Speaker labels you create
- Titles, tags, favorites, bookmarks, collections, and notes you add
None of this is transmitted off the device by VoxVault, except the private iCloud content you explicitly choose to sync, which goes to your own iCloud account.
No collection, no tracking
- VoxVault contains no analytics or tracking of any kind.
- VoxVault includes no analytics, tracking, or advertising SDKs. The on-device transcription and speaker components (WhisperKit and FluidAudio) are open-source and make no network requests.
- VoxVault does not create a user profile and does not require an account.
- Optional sync uses the private CloudKit database owned by your Apple ID. VoxVault's developer cannot browse its contents.
Trial and purchases
VoxVault uses Apple's StoreKit to activate the free trial, sell the optional Lifetime unlock, and restore access. Apple processes the transaction under your App Store account; VoxVault does not receive your payment-card details. The app reads Apple's verified entitlement status and stores only the minimum trial state locally in the device Keychain, including opaque recording identifiers needed to prevent the same recording from counting twice. That state contains no audio, titles, transcript text, summaries, speaker information, or advertising identifier, and it is never sent to a developer-operated server.
On-device processing
Transcription runs entirely on your device with WhisperKit — an open-source speech-to-text engine running OpenAI's Whisper model, bundled inside the app. Summarization uses Apple's on-device Foundation Models. Both run locally on your device. VoxVault runs no servers of its own; the only content that leaves your device is the private iCloud data you explicitly choose to sync, which goes to your own iCloud account (not to us). By default VoxVault processes each new recording automatically the moment you stop it (transcribe, identify speakers, tag, and summarize), all on-device, and each step can be turned off in Settings.
Permissions
- Microphone: requested only when you choose to start a recording, and required to capture that audio. Transcription needs no additional permission — WhisperKit works directly on the saved audio, on-device.
- Reminders: optional, requested only when you tap to add an action item to Reminders. VoxVault uses it to create the reminders you add (in a "VoxVault" list) and never reads or modifies your existing reminders. iOS has no write-only reminders permission, so the grant technically covers read/write, but the app only ever creates.
You can change these permissions at any time in the iOS Settings app.
Your control and deletion
- Deleting a recording removes its audio file, transcript, summary, speaker labels, and action items.
- "Delete All Recordings" permanently removes every recording and its files from the device.
- VoxVault runs no servers of its own, so deleting data in the app removes it from your device immediately. If you enabled iCloud sync, the delete also propagates to your private iCloud and your other devices. Copies inside previous device backups remain until those backups are replaced or expire; you can exclude VoxVault data from backups entirely in Settings.
Consent and recording responsibility
Recording laws vary by location. You are responsible for obtaining any consent required where you are. VoxVault shows a consent reminder and never hides that a recording is active; it does not support secret recording.
Speaker recognition (optional)
VoxVault includes optional speaker recognition:
- It is off by default. A master switch in Settings must be turned on before anything runs.
- Automatic speaker separation ("Speaker 1 / 2 / 3") and any named recognition run entirely on-device using models bundled in the app; they make no network requests.
- Named recognition requires you to explicitly enroll a voice, with per-profile enrollment consent. A voice profile is a small mathematical embedding derived from a few short samples, not stored audio, kept in the app only and never uploaded to any server.
- You can delete any voice profile, individually or all at once, or disable recognition entirely, at any time.
- Matches are probabilistic and shown honestly: below high confidence they appear as "Possible …?" or stay generic ("Speaker N"), and you can always correct a label. Your manual labels always take precedence.
This website
This policy covers the VoxVault app. The website you're reading (voxvault.vikw.ca) uses privacy-respecting, cookieless analytics via self-hosted Umami (pages, referrers, device and country only — no cookies, no cross-site tracking, no session recording, and no PII), and no other trackers.
Contact
For questions about this policy, contact: vikwapps@gmail.com.