Naming Conventions

The plugin reads two things off your folder structure: what kind of take this is, and which actor it belongs to. Get the layout right once and stereo takes find their own calibration and identity; get it wrong and takes queue as the wrong job type or block waiting for assets that are sitting right there.

The layout

Capture Manager ingest lays takes out one folder per actor, one folder per take:

/Game/.../Imports/
  Aaron/
    Aaron_Calib_Grid/       -> Stereo Calibration
    Aaron_Identity_DNA/     -> Identity Create
    Aaron_Take01/           -> Stereo performance
    Aaron_Take02/
  Bea/
    Bea_Calib_Grid/
    ...

The Imports/<Actor>/ segment is what the plugin parses to get an actor name. Calibration and identity discovery is scoped to that actor's folders, so Aaron's takes can never pick up Bea's calibration. A take that sits outside an Imports layout has no actor, and actor-scoped discovery cannot see it — mono and audio jobs do not care, but a stereo take there will never find its parents.

Name tokens decide the job type

New stereo takes are classified by matching their package path against wildcard patterns, case-insensitively. Calibration patterns win when both match:

*grid*, *calibration*, *calib* Stereo Calibration

*identity*, *dna* Identity Create

anything else Stereo performance

The patterns are editable under Ingest in Global Settings, and the whole behavior can be turned off — with auto-categorization disabled, every stereo take queues as a performance job and you set the two preflight types by hand.

Tokens are required only for stereo takes. Mono and audio takes need none — their type comes from the asset itself, and they never wait on a calibration or identity.

The failure this prevents

An actor's grid take must carry a calibration token and their identity take an identity token. Miss them and both fall through to ordinary performance jobs — which means no CameraCalibration and no MetaHuman Identity is ever produced, and every stereo performance for that actor blocks as MissingCalibration or MissingIdentity. The queue looks stuck for no visible reason; the cause is two folder names.

The mirror-image mistake is worse because it runs: those same tokens in a performance take's name or folder will misclassify it. A take called Aaron_Calibration_Test01 queues as a checkerboard solve and fails on footage that has no board in it. Keep grid, calib, identity, and dna out of names that are not those take types — including parent folder names, which are part of the matched path.

Fixing a bad layout

Renaming or moving Capture Data in the Content Browser is safe: the plugin's own rename leaves a redirector behind and updates any queued job pointing at the old path in place. It never overwrites an existing asset.

From a script or an MCP client, list_capture_data reports every take with the actor it resolves to — an empty actor is the tell that a take sits outside an Imports layout — and rename_capture_data performs the fix. See MCP Toolset.

After any rename, re-run validation so blocked jobs recompute against the new paths: validate_queue does it in one call.

The mha-organize-takes skill automates all of this — inventory, classification lint, and a proposed rename plan for raw footage on disk or Capture Data already in the project, read-only until you approve it. See Agent skills.

Naming raw footage before ingest

The layout is easiest to get right before the footage is in the project at all. If your stage lays footage out differently — an extra session level, say — the external-footage scan can be taught that layout with a schema file rather than reorganizing the disk. See Add Jobs dialog.