Caption spans in play order, covering every frame in [0, frameCount) exactly once, with adjacent spans naming different beats.
Output frames per second the spans are addressed in.
Total output frames covered: round(runtime * fps) for a whole render, or the window length after slicing.
Id of the sequence this sidecar captions.
The per-beat caption track of one sequence render: the machine-readable sidecar a guide-frame consumer reads next to the captured frames.
planCaptionSidecarproduces it for the whole render,sliceCaptionSidecarre-bases it to one chunk, andrenderCaptionSidecarserializes it.