Asoma Photo

Asoma Photo Privacy Policy

Last updated: 2026-11-13

This policy explains what data the Asoma Photo iOS app (“the App”) handles, where it is processed, where it is stored, and how to delete it.

Summary

1. What the App stores

All of the following lives in the App’s sandbox (a local database, image folders and the App’s preferences), except the home-screen widget data, which lives in a container shared only with the App’s own extensions (the widget, and the downloader iOS uses for the face models; see §5).

DataWhere it comes fromStored inLeaves the device?
ThumbnailsOne per imported photo, at most 1024 pixels on the long edgeSandbox image folderNo
Copies of originalsOnly for photos imported through the system picker while the App has no Photos accessSandbox image folderNo
Viewed originalsWhen you view a photo imported from Photos full screen, the original fetched from Photos is kept for next timeThe App’s cache folder; deleted automatically after 30 days unused, or least-recently-used first when the total exceeds 2 GBNo
Library referencesWhen importing from your Photos library, the photo’s Photos identifier and original file name; the original stays in PhotosLocal databaseNo
Capture time, device modelThe photo’s EXIFLocal databaseNo
LocationGPS coordinates from EXIF, and the country, city, district and time zone looked up from themLocal database; place names and time zones also in a local cacheAn approximate point is sent to Apple for the lookup, see §4
Face boxes, five landmarks, face cropsDetected on the deviceLocal database; crops in a sandbox image folderNo
Face identity vectorsA list of numbers computed on the device by a face model, used to tell whether two faces are the same personLocal databaseNo
Upper-body appearance vectorsComputed on the device from the upper body (mainly clothing), to help recognise people whose faces are unclearLocal databaseNo
Image analysisScene labels, body pose, sharpness, exposure, image fingerprint and whole-image features; for people you have named, whether a face is smiling and whether it wears glasses, sunglasses or a hat (used only to suggest their avatar). All computed on the deviceLocal databaseNo
People and relationshipsThe names, notes and categories you give, relationships between people, and your “not the same person” correctionsLocal databaseNo
Who is “me”, your avatarChosen by youPreferences; avatar in the sandboxNo
Home cityThe App suggests candidates from your photos’ locations; you choosePreferencesIf you use “current location”, see §4
Album titles, notes, favourites, pins, hidden peopleSet by youLocal database and preferencesNo
Home-screen widget dataUp to 9 of the day’s picked photos (reduced in size), a cut-out of the person in each, the place name and date, for the “Moment” widgetA container shared only with the App’s own extensionsNo
Performance logHow long each processing step took, photo counts, the file names photos have inside the App, App and iOS versionA text file in the sandbox, capped at 1 MBNo

Not collected: device identifiers, advertising identifiers, contacts, calendars, camera, microphone, health data, usage analytics, crash reports.

2. On-device analysis

The App analyses photos with the models and system features below. They all run on your device. The three face models are not part of the App download: after you install or update the App, iOS downloads them from Apple (see §4); the download contains only the model files, and no photo or other data of yours is sent.

Model / featurePurposeRuns on
RetinaFaceFinds faces and their five landmarksDevice (Core ML)
AuraFaceTurns a face into an identity vector so photos of the same person group togetherDevice (Core ML)
DINOv2Turns an upper body into an appearance vector, to help when a face is not clear enoughDevice (Core ML)
Apple VisionFace quality, scene classification, body pose, document detection, aesthetics and saliency, whole-image features, cutting a person out of a photo for the widget, telling whether a named person wears glasses, sunglasses or a hatDevice (iOS system framework)
Core ImageTelling whether a named person is smiling, to suggest their avatarDevice (iOS system framework)

No photo, face or other data is sent to any third-party or cloud AI service.

Not done: inferring age, gender, ethnicity or emotion. Whether a face is smiling or wears glasses, sunglasses or a hat is only used to suggest an avatar for a person you have named; it is not used for anything else and never leaves the device. Who “me” is, what people are called and how they relate to each other are set by you; the App does not guess them.

3. Permissions

PermissionWhen it is requestedWhat it is used for
PhotosOn first launch, and when you import photosReading the photos you choose to make thumbnails and analyse them; asking the system for the original when you view one full screen. With limited access, only the photos you allowed are read. Nothing in your Photos library is modified or deleted
Photos (add only)When you save a Day Album’s full-screen map as an imageAdding that one image to Photos. Your library is not read through this permission
Location (while using)Only when you tap “Use current location”One kilometre-accuracy reading to set your home city. No background location, no tracking
NotificationsOn first launchLocal notifications only: analysis progress and completion. No push server
Live ActivitiesWhile photos are being analysedProgress on the Lock Screen and Dynamic Island: photos done and the date being processed

You can turn off any permission at any time in the system Settings app.

4. Network access

The App does not connect to any server of its own. Only the following go online:

Apple’s privacy policy applies to the face model download, place names, time zones and maps; Cloudflare’s privacy policy applies to hosting the document pages.

Also, if iCloud Photos has “Optimise iPhone Storage” turned on, the system may download a photo (or a smaller version of it) from iCloud when the App asks for its original, or when it finishes analysing photos in the background. That happens between iOS and your iCloud; the App is not involved.

Apart from this the App makes no network requests: no analytics, no ads, no crash reporting, no sync.

5. Storage and backups

6. Deletion

The App offers the kinds of deletion below; viewed originals are also cleaned up automatically (after 30 days unused, or when they exceed 2 GB). Everything else stays until you delete the App.

ActionWhat is deletedWhat is kept
Deleting a Day AlbumAll of the App’s data for that day’s photos: thumbnails, copies of originals, viewed originals, face crops, library references, and those photos’ analysis (capture time, location, face boxes, face and upper-body vectors, which person they belong to, image analysis)The originals in your Photos library; people’s names, notes and relationships
Settings › Storage › deleting originals by albumOnly the App’s copies of originals, to free up spaceThumbnails and all analysis. When you later view such a photo full screen, the App asks Photos for the original to display it (the fetched original is kept as a viewed original); if Photos does not have it, or the App has no Photos access, only the thumbnail is shown
Deleting the AppEverything in the sandboxThe originals in your Photos library; existing device backups (see §5)

Hiding a person, removing a person’s identity and dissolving a relationship are not deletion: they change what is shown and how photos are grouped, and the related data stays until you delete the corresponding Day Album or the App.

7. Sharing and export

The App uploads nothing. Only this action, started by you, hands something to the system:

The App does not expose your data to Spotlight, Siri or other apps; the only thing outside the sandbox is the home-screen widget data (§5).

8. Third parties

9. Changes

Updates to this policy change the date at the top of this page. If an update materially changes how data is handled, the App will say so prominently.

10. Contact

Questions about this policy or your data: see the support page, https://asomaphoto.pages.dev/support/, which lists how to reach the developer.