AutoMovie API
    Preparing search index...

    Project one posed actor's joints to normalized 2D frame keypoints, the OpenPose-style conditioning automovie can emit exactly because it knows every bone's 3D position.

    Forward kinematics resolves the pose in rig space, the actor's staged node transform lifts each joint into scene world with the renderer's TRS order, and the live camera, sampled at time, projects it through an optional delivery crop. x grows rightward and y downward from the frame's top-left corner, both in [0, 1] inside the frame. A joint behind the camera, outside the clip range or outside the frame is still projected, never clamped, because a clamped point reads as a false edge keypoint, and is flagged inFrame: false. A requested joint the rig does not have yields no keypoint. The order follows joints. FK uses default engine joint axes and rest frames; coincidence with a renderer requires those same conventions, and these numerical outputs establish no observed rendered pixels.

    type IProps = {
        aspect?: number;
        camera: IAutoMovieCamera;
        cameraMotion?: IAutoMovieShot["cameraMotion"];
        crop?: IAutoMovieDeliveryCrop;
        joints?: readonly AutoMovieHumanoidBone[];
        node: { transform: IAutoMovieTransform };
        pose: IAutoMoviePose;
        skeleton: IAutoMovieSkeleton;
        time?: number;
    }
    Index
    aspect?: number

    Positive width/height ratio before delivery crop, default 16/9. Finiteness and positivity are caller premises; this function does not independently validate the ratio.

    The live camera.

    The camera's move; omit or pass null for a static camera.

    Optional normalized delivery crop; omitted projects the full raster. projectToNdc applies the crop after lens projection and delegates crop admission to deliveryCropNdc.

    joints?: readonly AutoMovieHumanoidBone[]

    Requested joint order, with duplicates preserved. Omitted uses DEFAULT_KEYPOINT_BONES; an empty list emits none and absent rig bones are skipped.

    node: { transform: IAutoMovieTransform }

    The actor's staged scene-node placement, in world space.

    Type Declaration

    • transform: IAutoMovieTransform

      Borrowed actor model-to-world TRS; translation is metres, scale is dimensionless and rotation uses the scene quaternion convention.

    The actor's pose at this instant.

    The actor's skeleton.

    time?: number

    Shot-local camera sample seconds, default zero. This does not resample pose; the caller must supply the actor pose for the same instant. A moving camera delegates time admission to sampleClip; a static camera does not read this value.