Create captions that carry non-speech meaning

Given [VIDEO TRANSCRIPT], [SCENE NOTES] and, if available, [TIMECODED VIDEO OR TRANSCRIPT], create a caption file plan that includes spoken words, speaker changes and only meaningful non-speech sounds. Keep reading speed suitable for [AUDIENCE], break at natural phrases and identify off-screen speech. Mark any inaudible word as uncertain; never guess from context. If timed media or a timecoded transcript is available, provide timestamped caption lines; otherwise give untimed caption text and a timing-to-verify checklist, never guessed timecodes. Also provide a sound-cue justification list and quality checks for contrast, safe area and mobile playback. Distinguish literal captions from edited subtitles. Do not add emotional labels like 'angry' unless performance supports them.