What Memoria does, feature by feature. Everything runs on your device.
Across text in images, words spoken in videos, object labels, person names, places and camera metadata.
As you type, Memoria suggests terms from your own library's vocabulary (never from a remote dictionary).
Narrow results by photo, video, indexed, not-indexed-yet, needs-reindex, or 'no text found'.
Memoria reads the text in every photo, using your device's native engines.
Key frames are sampled from videos and processed for text, which is useful for slides, conference footage and on-screen captions.
Three modes (Fast, Balanced and Quality) depending on whether you favor speed or detail.
English, French, German, Spanish, Italian, Portuguese, Japanese.
Several model sizes run entirely on your phone, letting you choose the balance between accuracy and disk space.
Whisper detects and transcribes spoken language automatically.
A lighter alternative, with no download, when you'd rather not install a Whisper model.
Faces are detected and recognized entirely on your device.
Memoria groups faces into people. You name and merge the groups; Memoria never sees them.
Tune the grouping tightness: looser to merge variants, tighter for finer distinction.
Memoria identifies the objects in your images, with labels in your language.
GPS metadata becomes searchable place names ('Berlin', 'kitchen', 'beach') without sharing your location with anyone.
Make, model and focal length are extracted and searchable.
From 1 to 5 columns; the grid stays smooth on the largest libraries.
Zoom on images, integrated video player, sliding info drawer with extracted text, objects, faces and place.
Indexing continues even while the app is closed.
Automatic pause on low battery; optional 'only while charging' mode.
An activity bar at the bottom of the screen shows real numbers: 12,403 of 38,210. No vague spinners.