Plan Silence Cuts with Source-to-Output Maps

What it does

Analyze the first audio stream with silencedetect and propose quiet-region removals using a dB threshold, minimum silence duration and preserved boundary padding. Write raw quiet intervals, proposed cuts and a kept source-to-output timeline in JSON, plus the kept timeline CSV.

How to use

Run --dry-run to analyze without writing. Remove it to save a new report JSON and sibling CSV. Review every proposed cut against the visuals. Only an explicit --export new-cut.mp4 creates an edited version. Entirely quiet, no-audio, empty/no-cut and over-200-range timelines are refused for export.

Requirements

Python 3.10 or newer and separately installed FFmpeg/ffprobe. No Python packages or binaries are bundled. FFmpeg must provide libx264 and AAC for reencoded exports.

Caveats

Silent visuals may matter. This is amplitude-based detection, not speech recognition or semantic editing. Export reencodes the first video/audio streams; other streams are omitted. JSON/CSV mappings are proposed continuous timings, not frame-exact guarantees. All output paths must be new. It does not download arbitrary YouTube URLs.