Face Data
How Prism Studio handles face data.
Prism Studio is a portrait and face-effects app, so some user-selected images may contain a face and some
features use on-device face detection or landmark tracking to align effects. This section explains exactly what
face data Prism Studio handles, how it is used, when it is shared, and how long it is retained.
Face data details
-
What face data Prism Studio collects
Prism Studio may collect user-selected photos or video frames that contain a face. For on-device preview and
recommendation features, Prism Studio may also derive face-related measurements such as face presence, face
count, face bounding boxes, approximate face contour points, and landmark positions for features like the eyes,
nose, mouth, chin, and top-of-head placement.
-
How Prism Studio uses face data
Face data is used only to provide the features the user requests inside the app, including aligning live face
effects, identifying whether an imported image is a portrait or person image, recommending compatible Morphs,
and generating or analyzing portrait-based results when the user explicitly chooses a cloud AI feature.
-
No identity recognition or face-template building
Prism Studio does not use face data to identify a person, authenticate a person, build a biometric profile, or
create a reusable facial-recognition template across users.
-
Sharing and storage
On-device face landmarks, face rectangles, and similar live tracking measurements stay on the device for local
preview and are not shared with third parties. If the user asks Prism Studio to run a cloud AI scan,
generation, or finishing workflow, the selected image itself may be sent through Prism Studio's backend to the
provider used for that request, including OpenAI for image analysis and planning and Replicate-connected
providers for image or video generation. Imported media, saved results, and saved Morph data are stored locally
on the user's device unless the user submits them to a cloud AI workflow.
-
Retention
Live on-device face tracking measurements are processed in memory for the active session and are not stored as a
long-term biometric template. Imported media and generated results stay on the user's device until the user
deletes them. Derived remote image-scan results that can include portrait classification, face count, and face
regions may be cached on-device for up to 7 days to avoid repeated scans. Uploaded cloud AI requests are
retained only for the time reasonably needed to process the requested feature and return the result.