Your data stays with you. Period.
KontexVoX stores your photos and descriptions on your device. The app sends neither your photos, nor your audio, nor your descriptions to our servers. No ad tracking, no third-party analytics.
If you use iCloud Photos sync, your enriched photos sync like any other photo. That's normal iOS behavior. But us, the KontexVoX app, we send nothing. The distinction matters: iCloud is Apple, not us. We don't control it, and enabling it is your choice.
Transcription is performed by iOS's native speech recognition, locally on your phone. No audio is sent to a third-party server. It may produce a few small imperfections, but that's the price of privacy. And it improves with every iOS update.
We could have used a cloud transcription like OpenAI's Whisper, more accurate, but it would have required tokens and we couldn't have guaranteed that your data stays your data. We chose local.
We compared 100% local transcription (airplane mode, no network) and online, on the same sentences. Result: equivalent quality — local even punctuates better more often than not.
A home test, in full transparency: two sentences dictated identically — one realistic, one deliberately "tricky". Not a benchmark, just what we saw with our own eyes. (Test run in French, our first-language market — the takeaway holds in any language.)
« …ranger dans le bureau, …deux verres sans pieds… » — the word "vert" (green) is dropped, punctuation is weak.
« …ranger dans le bureau vert… À côté, deux vers 100 pieds… » — keeps "vert" (green), punctuates better.
Codes, HEIC format, numbers and units (1.2 MB, ISO 800, f/1.6) and jargon ("durite"): correct in both modes. Punctuation and sentence splitting: better on local. Homophones (ver / verre / vert / vers, comte / conte / compte) fail identically in both modes — that's a limit of Apple's engine, not of local mode.
Bottom line: neck and neck. Local is at least as good, and cleaner on punctuation.
Conclusion: choosing 100% local costs nothing in accuracy. So we chose it, all the time. Your data stays with you.
The transcription of your voice is written into each photo's EXIF metadata, as text. No audio file attached. These metadata travel with the photo file. If you share the photo via AirDrop or the Files app, the metadata is preserved. If you share via WhatsApp or Telegram in image mode, the metadata may be stripped by the third-party app.
If you allow location access, GPS coordinates are added to the EXIF. It's optional: you can refuse in iOS settings at any time. Evidence mode does not yet include geolocation; it's planned for a future version.
All your data lives in your photos. Deleting a photo deletes its context. There's no server to clean up.
Creating an account is optional: it only serves to raise your daily quota. In that case, your email address is stored with our authentication provider (Supabase). It's the only personal data we keep, and you can request its deletion at any time.
If you go Premium, the purchase is handled by the App Store and our subscription provider (RevenueCat) processes your purchase status. Neither receives your photos, your audio, or your descriptions. No data is used for ad tracking.
For any question: support@kontexvox.com