Caption Timing Auditor

Scenario

Use after rendering or caption generation when subtitles must stay aligned with word-level timestamps.

Input and output

Input contains word timestamps and caption cues with start, end, and text. Output reports drift, missing words, overlapping cues, and caption_hash.

Execution and recovery

Run python scripts/audit.py examples/input.json. Invalid boundaries and unsupported cue order return visible errors. Adjust the caption file and rerun; no video is rendered.

Privacy boundary and acceptance

Use redacted words and media references only. The sample must produce a PASS or REVIEW decision and a non-empty hash.