Id of the scene camera that is live for this shot.
OptionalcameraSwept-clearance reports for hero and alternate takes declaring physical envelopes, one per such camera. The performing producer checks the runtime revision; the stored report has no revision pair for an artifact reader to re-establish freshness independently.
Omission is admitted only when no delivered camera declares an envelope. A present list must carry clear reports with empty findings; blocked or stale evaluation refuses publication by the performing producer.
OptionalcameraThe camera's directorial intent per frame span: framing, move, the
resolved focus point, and the lens intent. Structural guide metadata a
diffusion or render host reads beside cameraMotion, as it reads events.
The deterministic camera solve never consumes it. Absent means legacy; an
empty array means the shot was assembled with no frame actions.
The camera's move for this shot: a clip of the camera node's transform
tracks. null for a locked-off (static) camera.
OptionalcoverageThe alternate camera takes covering the same beat: one staged camera per
additional angle, each with its own compiled move and intent. The hero
take stays the singular camera and cameraMotion every consumer already
reads; coverage rides beside it as structural guide metadata a diffusion
or render host uses to render the beat from the other staged angles. The
cut is untouched: coverage takes are alternates of this shot, never
separate timeline entries. Absent means legacy or single-camera data; an
empty array means the shot was assembled with one camera.
Finite positive shot length in seconds (local time origin = 0).
OptionaleditorialAuthored shot-local annotations in authored order. Omit when no marker list was supplied.
OptionaleventsComputed or scripted interactions on this shot's local clock, ordered by
time, then kind, then id. performShot emits this for contacts, hits,
falls, and attach handoffs so downstream tools can follow the same timing
the engine used. Absent means legacy or no-event data; an empty array
means the shot was assembled and had none.
Stable, non-blank id.
OptionallightClips changing this shot's lighting over its local clock: a candle going
out, a sunset. Each track addresses one staged light by the pointer
/lights/<id>/<property>, where the property is intensity, color,
range, coneAngle, position or rotation, and no two tracks across
the list may address the same light property. Absent means legacy; an
empty array means constant light.
Human or LLM readable name, or null when unnamed.
Node-transform clips for non-skeletal scene objects the shot animates (a
launched projectile's baked flight, a prop carried along a path, a door
panel or prop joint the source authored), each an ordinary
IAutoMovieClip with a unique id. Distinct from performances
(skeletal pose motions played through a rig) and from cameraMotion (the
one live camera): an object has no skeleton, so it moves the way the camera
does, by a clip of transform tracks. Empty when the shot animates no such
object.
Per scene-node performances for this shot, at most one per node.
OptionalpolygonPolygon body changes in source-shot time. Each subject and bone pair may appear once.
OptionalpracticalMaterial tracks for named parts of placed props. At most one per node and part. Omission or an empty list leaves the staged appearance in force.
Id of the scene (placed models, lights, cameras) this shot renders.
A shot: one continuous take (a scene, the camera that frames it, the camera's move, and what every placed node performs) over a local time range. This is the unit an LLM renders ("render this shot") and the rung above a single clip on the road to assembling a film from objects and motion.
Time is local to the shot (origin 0, seconds); a IAutoMovieSequence composes shots into a global timeline. The camera is a scene node, so its move is an ordinary IAutoMovieClip of node transform tracks, with no special camera-animation concept.
performShotassembles the shot andvalidateShotArtifactis its structural contract, which every producer and consumer shares. A clip field admits only the channels its own applier writes:cameraMotionandobjectMotionstake node channels, andlightMotionstakes light pointers, so a pointer such as a camera's field of view is refused rather than stored and ignored.