# Clip report

- source: `sample.mp4` (40.6 s)
- transcript: faster-whisper `small`, CPU int8, 9 segments / 97 words
- renderer: local ffmpeg (libx264/aac), burned-in ASS captions
- selection: content analysis: audio energy, scene cuts, transcript salience

## Clip 1 — 0:00:10.92–0:00:17.78 (6.9 s)
- why: audio energy 0.64 of the 95th percentile; selection method: content analysis (no retention data). Transcript: "The second signal is a scene change. When the picture cuts, attention resets, and a clip can start there."
- formats: clip01_vertical.mp4
- boundary: sentence boundary (whisper segment)

## Clip 2 — 0:00:18.76–0:00:26.20 (7.4 s)
- why: audio energy 0.65 of the 95th percentile; selection method: content analysis (no retention data). Transcript: "The third signal is the transcript itself. Questions, answers, and repeated keywords show where the topic turns."
- formats: clip02_vertical.mp4
- boundary: sentence boundary (whisper segment)

## Deliverables

- `clip01_vertical.mp4`
- `clip02_vertical.mp4`