Read the project-level audio timeline.

Returns the multi-track audio timeline: music layers with their clips — a clip may span several scenes and crossfade into the next — plus sound-effect cues placed at an absolute offset within a scene. Every sound-effect cue in the answer carries how loud it really is: assetLufs (its file: a bed's body, the median of its sounding 400 ms windows; a spot's loudest 400 ms), normGainDb (what the mix lifts or lowers the file by before its level: a bed is levelled to -17 LUFS, a spot to -18, never past -1.5 dBTP or more than +30 dB), mixLufs (where it sits in the mix at its level; the sound map reads it against its scene's own mix, and the sound map itself lists a capture beat's own click sounds with their loudness too), a bed's continuity (the share of its span within 10 dB of its body) and loudnessFlag (too_quiet: its file came out too quiet to bring up, make it again; clipped; gappy: a bed that drops out or fades, make it again as one continuous sound; unmeasured: measured at the next stitch). Also returns the scene axis (label, duration, start image) that clips are positioned against. This is the through-composed alternative to per-scene beds, which are generated independently and therefore share no tempo, key or phrase. Scope: story:read.