MarkVox Privacy Policy
Last updated: 2026-06-21 Last checked: 2026-06-26
Status: publication-candidate draft for counsel review. Do not publish until approved.
MarkVox helps you capture highlights from books, articles, screens, and other reading material by using OCR, short voice anchors, source metadata, notes, folders, and exports.
Our Privacy Commitments
voice recordings, transcripts, or exports for advertising.
anchors, transcripts, source identifiers, titles, authors, URLs, file paths, photos, or audio in diagnostics.
shutter uses local audio-level detection only and deletes temporary recorder files best-effort after the local trigger flow ends.
- We do not sell your personal information.
- We do not use your highlights, notes, OCR text, source metadata, photos,
- We do not intentionally include OCR text, highlight text, notes, spoken
- We delete temporary voice audio after the transcription attempt. Sound
- We provide in-app ways to export account data and request account deletion.
Data We Process
| Data category | Examples | Why we use it | Storage posture |
|---|---|---|---|
| Account data | Email, user id, account timestamps | Sign-in, account recovery, account management | Stored while the account exists |
| Highlight content | Highlight text, notes, page number, folders, source link | Core library, editing, search, export | Stored until deleted or account deletion |
| Source metadata | Titles, authors, identifiers, URLs, publication details | Citation, export, source library | Stored until edited/deleted or account deletion |
| OCR text and evidence | Raw/cleaned OCR, quality metadata, page linkage | Matching spoken/typed anchors to captured text and quality review | Linked OCR may be stored; unsaved orphan captures are eligible for cleanup |
| Photos/images | Camera/library page images and temporary draft images | OCR, confirmation, and draft resume | Sent for OCR; draft images may be temporary; saved highlights do not rely on a user-facing photo archive |
| Voice recordings | Short anchor/source/folder recordings | Transcription into text commands | Deleted after transcription attempt |
| Local sound trigger | Sound shutter metering after the user explicitly arms it | Local camera trigger | Not uploaded or transcribed; temporary recorder file deleted best-effort |
| Voice transcripts | Spoken anchors, source commands, folder voice titles | Matching, source entry, folder filtering | Anchor/source-command transcripts are transient; folder voice title transcripts may be stored |
| Preferences and consent | Text size, high contrast, privacy choices, voice disclosure | Accessibility and app behavior | Stored locally and/or account-backed |
| Match feedback | Suggested/final highlight text classification, verdicts, correction signals | Quality analysis and product reliability | Stored without photos or audio |
| Diagnostics | Sanitized error area/action/status | Reliability and crash triage | Production sink is not currently wired |
| Analytics | Consent-gated interaction aggregates | Product improvement | Production sink is not currently wired |
Photos, OCR, And Captures
When you photograph or import a page, MarkVox sends the image to OCR providers so it can extract text and match your spoken or typed anchors. MarkVox may store OCR text and quality metadata when needed to support saved highlights, draft captures, reliability analysis, and account data rights.
Unsaved page captures that are not linked to a saved highlight are eligible for cleanup after 7 days.
Voice Data
MarkVox uses short voice recordings to turn your spoken commands into text for the feature you requested. Current transcription features include highlight anchors, source identifier voice entry, and folder voice commands/titles.
Sound shutter is different: after you explicitly arm the camera, MarkVox uses local audio-level metering to detect a clear sound and take a photo. It is not sent to Whisper, and no transcript is created.
MarkVox deletes voice audio after the transcription attempt. MarkVox does not use current voice recordings for advertising, sale, or voice training.
Voice-training or speaker-recognition features are not part of the current product. If added later, they will require separate consent, retention wording, withdrawal controls, deletion controls, and counsel review.
Source Lookup
If you ask MarkVox to look up source metadata, the app may send identifiers or title candidates to services such as Google Books, Crossref, PubMed, or arXiv. These requests are used to help fill citation/source fields for your review.
Account Export
You can export account data from Profile -> Data & Privacy. The current export includes account metadata, saved highlights, notes, folders, sources, selected preferences/consent state, match feedback, and linked OCR quality metadata.
The current export excludes raw/full OCR text, cleaned OCR text, page image URLs, local draft image URIs, draft OCR text, and audio files by default.
Account Deletion
You can request account deletion from Profile -> Data & Privacy. The current process records a deletion request for manual review and completion.
When processed, account deletion should remove or anonymize account data, highlights, notes, folders, source metadata, match feedback, linked OCR evidence, and preference/consent records, unless limited retention is required for legal, security, abuse-prevention, or accounting reasons.
Draft processing target: review requests within 5 business days and complete or respond within 30 calendar days. This target is pending counsel approval.
Processors
Current or planned processors may include:
disclosed.
- Supabase for authentication and account/content storage.
- Google Cloud Vision and limited Google Document AI fallback for OCR.
- OpenAI Whisper for transcription. Sound shutter does not use Whisper.
- Google Books, Crossref, PubMed, and arXiv for source metadata lookup.
- Expo/EAS for app development and distribution tooling.
- Future diagnostics, analytics, or billing processors only if wired and
Processor contracts, data-processing agreements, regional settings, and vendor retention controls must be confirmed before public launch.
Children
MarkVox is not intended for children. MarkVox should not be used by children until the app has an approved age gate and child/privacy posture.
Your Rights And Choices
Depending on your location, you may have rights to access, correct, export, delete, object to, restrict, or withdraw consent for certain processing. You can start with in-app controls in Profile -> Data & Privacy.
Privacy contact: privacy@markvox.app
Support contact: hello@markvox.app