AGENT SKILL

pause-detector

Detect long low-energy pauses using a local dynamic threshold on precomputed energy data (robust to volume changes across a lecture).

License 许可
Apache-2.0
最近更新
2026-11-14
Stars
★ 0
Forks
1

01 前言 Front Matter

name
pause-detector
description
Detect long low-energy pauses using a local dynamic threshold on precomputed energy data (robust to volume changes across a lecture).

02 资源文件 1

03 SKILL.md 正文

Steps

  1. Decide where to begin pause detection:
    • Use --start-time <opening_end> to skip the non-teaching opening (where <opening_end> comes from silence-detector), otherwise use 0.
  2. Run the bundled pause detector from the workspace:
    python3 pause-detector/scripts/detect_pauses.py \
      --energies /path/to/energies.json \
      --output /path/to/pauses.json \
      --start-time <seconds> \
      --threshold-ratio 0.5 \
      --min-duration 2 \
      --window-size 30
    
  3. Tune parameters (match the script’s condition energy < local_avg * threshold_ratio):
    • threshold_ratio: a second is a “pause” when its energy is below a fraction of the local average.
      Increase the ratio (e.g., 0.5 → 0.7) ⇒ threshold rises ⇒ more seconds/segments detected as pauses (more aggressive removal).
      Decrease the ratio (e.g., 0.5 → 0.3) ⇒ threshold drops ⇒ fewer pauses detected (more conservative, fewer false positives).
    • min_duration: increase ⇒ only longer pauses are removed; decrease ⇒ more (shorter) pauses are removed.
  4. Confirm segments look valid: numeric start < end, duration == end - start, and times are within the video duration.

Expected Result

A pauses.json containing a segments list of disjoint low-energy intervals (seconds) meeting the minimum duration.

04 同仓库更多 Skills