AutoMovie API
    Preparing search index...

    Function reviewVisualRead

    • Engine-computed visual-read metrics for one compiled shot, filed as advisory tier: "visual" review notes after any film-grammar findings passed in. They never fail a gate; they give the review backlog measurable facts about whether the action reads.

      Every metric reads a performer's world root: its performance's motion sampled at max(0, t - startOffset) and folded through its staged node. A performance whose node or motion is missing, or that plays no motion, is skipped.

      • Subject in frame: sampled at sampleRate from 0 to the shot's end, a performer whose root leaves the live camera's frustum, behind the near plane, past the far plane, or outside the vertical field of view and the aspect-derived horizontal one, gets one note at the first such time.
      • Silhouette separation: at the same times, two performers whose angular separation from the camera is smaller than the sum of their angular radii, atan(silhouetteRadius / distance) each, get one note at the first such time; a pair with either root nearer than the near plane at that time is left to the framing metric.
      • Contact connection: a hit or contact event with a point and a target whose performer's root lies farther than contactRadius from the point at the event time gets one note.

      The first two metrics need a live camera with a finite field of view in (0, 180) degrees and far > near; without one they are skipped. A root is a point, so a subject whose body is in frame while its root is not is reported; that approximation holds for all three metrics.

      Parameters

      Returns IAutoMovieReviewNote[]

      The grammar notes followed by the visual-read notes.