Run local source analysis. Preview results; apply=true saves a new timeline revision.
masks options: mode (chroma, luma, alpha, grabcut, refine), key, tolerance, softness, rect, feather;
mode refine (an existing matte): grow (px, - shrinks), feather, fill_holes, smooth (frames), keep_largest. faces: min_size,
detector (auto: the bundled YuNet model with eyes, nose tip and mouth corners; cascade: OpenCV cascades, eyes
only), cascade (a classifier file, cascade detector). interpolate: multiplier. retime (slow motion as a plate with
one new picture per output frame): start, seconds (clip source interval), output_frames, fps (default the
canvas fps), take_start/take_seconds (one take; cuts inside are refused), deduplicate (collapse baked repeats,
run-centre timing), duplicate_threshold; returns a new clip to add and use at speed 1 from src 0 (apply is not
available for retime). No models are automatically downloaded.
(project_id: str, clip_id: str, operation: Literal['masks', 'faces', 'interpolate', 'retime'], options: dict | None = None, apply: bool = False) -> dict
By rk0neki·Created about a month ago·Updated 16 days ago· 82