AI features and your permission
AI scans send your selected photos (including faces), makeup references, public video links, selected frames, relevant extracted audio and your beauty profile to Aphrodite, OpenRouter and the selected AI model provider only to create the result you request. The sign-in page clearly discloses this sharing before its account-creation and email, Google or Apple sign-in actions. Completing one of those actions records your authorization for this account on this device; merely restoring a session or accepting a general policy link does not grant permission. Sign-in does not upload media, start a scan or use Credits. Permission is remembered and is not requested on each scanner screen. You can also enable or revoke it here. Materially changed data use requires renewed permission. Revocation stops new submissions, not an already submitted job; history or account deletion removes stored results. The onboarding walkthrough uses a local animation and locked sample content; no AI analysis or Credit debit takes place. Public-link previews may use a metadata lookup, but your chosen photo/video file is not uploaded by that walkthrough. After purchase, a separate explicitly priced scan is required to produce personal results.
Face data we collect
Face data includes the photos or selfies you deliberately choose, selected video frames containing faces, submitted makeup-reference portraits, and generated makeup previews. Depending on the requested feature, processing derives temporary facial landmarks, face counts, eye/lip/brow/cheek/complexion regions and geometric proportions; cosmetic observations such as apparent skin colour, undertone, contrast, eye opening or lip balance; and your requested cosmetic guidance. Preview results can store normalised face-region crop rectangles so the app can display close-ups. Saved beauty-profile choices may include a sampled skin-colour hex value, estimated depth and undertone, and face or eye shape you enter yourself. These are cosmetic observations, not calibrated measurements, diagnoses or conclusions about identity. Aphrodite does not obtain Face ID templates or use the TrueDepth/ARKit face-tracking APIs.
Face data: purpose and limits
We use face data to locate cosmetic regions, sample colours, suggest palettes and makeup placement, adapt a chosen style to visible features and your preferences, produce the makeup preview you request, and check for missing faces, unchanged images or obvious structural drift in generated previews. The basic colour-sampling calculation runs on your device; its photo is not uploaded for that calculation. A separately requested AI analysis or preview sends the selected resized media for server and AI processing. Temporary landmark coordinates are used in processing memory rather than enrolled in a face-recognition database. Coarse cosmetic descriptors and preview crop rectangles may be retained with results or job metadata. We do not create reusable faceprints, identify or authenticate people from their faces, match faces across accounts or a face database, or use face data for advertising, sale or training Aphrodite models. No additional use is planned beyond the requested cosmetic features and operating and securing them. A materially different use requires renewed disclosure and permission.
Face data: sharing and storage
Yes, requested AI features share face-containing images with third-party processors. Google Firebase/Google Cloud hosts authentication, private Firestore queue/results records and private Cloud Storage preview images. Our production Firestore database and preview-image bucket are in asia-southeast1 (Singapore). OpenRouter receives selected resized images and relevant cosmetic context and routes them to the serving AI provider. Configured routes include DeepInfra for Ling vision analysis, Alibaba/Alibaba Cloud for Qwen analysis or image generation, and Meta for Muse image generation, subject to route availability and the feature requested. These providers process remotely and are not guaranteed to stay in Singapore. The images are not sent to RevenueCat for billing or to retailers for product discovery. RevenueCat receives account and purchase information, not face images. Backend images use authenticated owner-scoped access; the preview bucket blocks public access and we do not intentionally publish selfies. Temporary facial-landmark processing also runs in our backend; it does not require sending those landmark arrays to an identity service.
Face data: retention
Different face-data types have different retention periods. (1) Basic colour-sampling photos and sampled colours are processed on-device; saved local cosmetic settings and thumbnails remain until removed or app data is cleared. Selecting that same photo for a later AI feature is a separate upload. (2) Queued resized analysis images are removed on worker completion or failure; otherwise their queue documents expire after 15 minutes and are physically removed by asynchronous TTL cleanup. (3) Analysis job and insight records, including stored preview crop rectangles, expire from normal reads after 30 days and have asynchronous Firestore TTL cleanup. (4) Glow-up selfie, reference and generated-preview image objects become eligible for automatic deletion at age 30 days; lifecycle execution is asynchronous. The preview bucket currently has no soft-delete recovery window and no configured object-versioning retention. (5) Separate Glow-up job records, including coarse cosmetic appearance/profile descriptors, and user-saved result metadata are not all covered by those automatic deletion policies. They can remain until server-history, saved-item or account deletion, even after their normal display window or image expiry. (6) Firestore recovery copies of deleted records can remain within the configured seven-day point-in-time recovery window. (7) OpenRouter and serving AI providers have separate processing, logging, safety and legal-retention rules. Our chat-analysis routing denies endpoints that declare training/data-collection use; this is not a zero-retention guarantee and does not establish an identical policy for image-generation endpoints. We cannot promise one fixed upstream erasure period or that deleting Aphrodite history deletes processor copies. OpenRouter describes optional content logging and input/output use, abuse-screening exceptions and provider-specific retention in its current Data Collection and Provider Logging documentation. Some provider retention periods are not publicly fixed. We do not represent these external services as universally zero-retention or guarantee no provider-side training where the applicable endpoint terms do not establish it.
Face data: deletion and your choices
Choose only media you have permission to process. Choosing or cancelling a photo, viewing a screen or signing in does not itself start an AI scan. AI sharing is disclosed at sign-in and can be enabled or revoked in Privacy Policy; revocation blocks new submissions but cannot retract a job already submitted. In the app, open Profile, then Privacy & data, then Delete server analysis history to remove owned server analysis records, Glow-up job records and preview images after active work finishes. User-saved looks are separate: remove them individually, or delete your account to remove owned saved cloud data as well. Profile > Delete account requires recent authentication and deletes owned cloud data and images, subject to the recovery window above; minimal billing/anti-duplicate-purchase records are retained separately and are not a face-recognition database. Deleting server history does not erase independent on-device thumbnails, guest saves or saved cosmetic settings; remove those separately or clear app data. Deleting an account does not cancel an Apple subscription. Processor-held safety, legal or other retained copies follow the processor terms; no immediate universal third-party erasure is promised. Contact support@aphroditeapp.site for privacy or deletion assistance.
Purchases and account credentials
Apple processes App Store payments; Aphrodite does not receive card or bank details. RevenueCat validates purchases using your Firebase account ID and store transaction information. The server stores purchase claims and a Credit ledger in Firebase to prevent duplicate grants and reconcile balances. Firebase login credentials are stored in the device protected credential store, not ordinary app preferences.
What we process
When you submit a link, we process its public metadata, captions and selected frames. When captions are missing, we may send temporary extracted audio through OpenRouter for speech-to-text to inform product and routine notes. When you submit a photo, the app resizes it for analysis. We process your input, derived product candidates and routine, account ID if signed in, and technical request data needed to operate the service.
Why and with whom
We use submitted media only to provide the cosmetic analysis or preview you request and to operate and secure that service. Selected media and relevant cosmetic context are sent through OpenRouter to the serving AI provider. The face-data sections below identify the processors and explain their separate retention practices. Aphrodite does not train its own models on your face data. Product reference information may be obtained from public catalogs, including Open Beauty Facts; product discovery does not require sharing your selfie with retailers.
Glow-up makeup previews
For a requested makeup preview, Aphrodite sends a resized selfie and makeup reference through OpenRouter to an image-generation provider. A submitted custom reference may also be sent for makeup analysis. We store the resized selfie, reference and generated preview in private Google Cloud Storage so you can reopen or request regeneration of your preview. The glowup/ object prefix has an age-based 30-day deletion lifecycle; deletion is asynchronous, not exact-clock erasure. Requests carry inline images, not public selfie URLs. Descriptive job metadata is separate from image objects and can remain until server-history or account deletion. See Face data: retention and Face data: deletion and your choices for the complete distinction.
Storage
The production queue stores resized analysis photos or selected frames in server-only Firestore scanPayloads documents. Workers delete those payloads on completion or failure. Unprocessed payloads have a 15-minute expiry and an active asynchronous Firestore TTL deletion policy. Analysis job and insight records have a 30-day read expiry and active TTL policies. Separate Glow-up job metadata can remain until server-history or account deletion. Successful video transcripts expire from cache reads after 24 hours, but their collection has no active physical TTL policy; stored transcripts can remain until history or account deletion. Temporary extracted audio is removed after source processing; full source-video files are not intentionally retained. Deleted Firestore data may remain in the configured seven-day point-in-time recovery window. Signed-in saved looks and named collections sync through Firestore until removed. Device-only thumbnails and local cosmetic settings are separate from server history. Preview-guest collections stay on this device.
Your choices
You can use Aphrodite without an account. Remove saved looks individually, delete server analysis history here, or delete your account from Profile. Account deletion removes your saved cloud data and derived history; minimal purchase records are retained to prevent duplicate grants and reconcile refunds. Deleting your account does not cancel your Apple subscription. Device-only thumbnails and guest saves are removed when you clear app data.
Security and third parties
We restrict backend requests to supported source hosts and verify generated shopping destinations separately. External retailer and content sites have their own privacy practices. No third-party analytics SDK is enabled in this preview.
Changes
We may update this policy when storage, providers or features change. Material changes should be described before a new data use begins.
AI processor policies
OpenRouter: Privacy Policy, Data Collection, and Provider Logging and retention. These documents explain external processing and link provider-specific terms; they do not override the Aphrodite retention distinctions above.
Contact
Questions about privacy: support@aphroditeapp.site.