Face Labeling
Name a face once — every scene they're in becomes searchable by name.
You remember who was in the shot, not what the file was called. Orcah detects and clusters every face while it indexes, then hands them to you unnamed. Put a name to a face once and that name follows the person across the whole library — every scene they appear in becomes something you can search for, on its own or combined with what was said.
Getting started with Face Labeling
- 1.
Point Orcah Studio at a folder
Faces are detected and grouped as part of the same indexing pass as scenes, transcription and objects. There is no separate step, and no photo leaves your Mac.
- 2.
Name the faces once
The Unnamed row holds every face that has no match yet. Select one — or a few of the same person — hit Label and type the name. Dragging a face onto an existing person works too.
- 3.
Let the name propagate
Labeling runs a pass over the library: every scene that person appears in picks up the name, and Select similar finds the faces that are probably them but haven't matched yet.
What it can do
- Automatic detection: every face in every scene is found and clustered during indexing — no manual tagging to get started.
- Unnamed, not guessed: faces arrive as an unnamed cluster rather than a wrong name, and stay that way until you label them.
- Batch labeling: select several faces of the same person and name them in one action.
- Select similar: pull the faces that look like the selection, so stragglers and odd angles get swept into the right person.
- Library-wide propagation: a name given once applies to every appearance, past and future.
- Searchable by person: combine a name with a topic, a place or a look — find anyone, anywhere in the library.
- Runs on-device: faces are detected and matched on your Mac, which keeps NDA footage and unreleased cuts where they belong.
Pro Tip: label the faces with the most appearances first — each one names hundreds of scenes at once, so the library becomes searchable by person fastest.
Questions worth asking
- What happens to faces it can't identify?
- They sit in the Unnamed row rather than being guessed at. You see an unlabelled face with its appearance count until you give it a name.
- Can I search for a person and what they said at the same time?
- Yes — that is the point of naming faces. Once a person is labelled you can ask for the moments where they appear, combine it with a transcript search, and the agent intersects the two for you.
- Are the faces uploaded anywhere to be recognized?
- No. Detection and matching models download once, then everything runs on your Mac. A face is only ever compared against your own library.
Find the moment. Not the file.
Download immediately, models included. Your footage stays on your Mac.