extract-video-audio

name	extract-video-audio
description	Use when extracting audio tracks from hse_ts_course lecture or seminar video files into committed content-level audio/raw folders for transcription pipelines.

Extract Video Audio

This skill extracts audio from course videos into audio/raw directories inside the content folder being transcribed.

Directory Contract

Choose the semantic content folder first:

<year>/lectures/videos/  ->  <year>/lectures/audio/raw/
<year>/seminars/videos/  ->  <year>/seminars/audio/raw/

Legacy <year>/videos/ inputs are still accepted for lectures, but new outputs should be written under <year>/lectures/audio/raw/.

Workflow

1. Locate the repository root and the skill directory:

REPO_DIR="$(git rev-parse --show-toplevel)"
SKILL_DIR="$REPO_DIR/.codex/skills/extract-video-audio"

2. Preview what will be extracted before writing files:

uv run python "$SKILL_DIR/scripts/extract_audio.py" \
  --repo "$REPO_DIR" \
  --year-dir 2026-spring \
  --kind lectures \
  --dry-run

3. Extract audio. The script creates audio/raw automatically and writes one audio file per input video:

uv run python "$SKILL_DIR/scripts/extract_audio.py" \
  --repo "$REPO_DIR" \
  --year-dir 2026-spring \
  --kind lectures

For seminars, use --kind seminars.

4. Check the generated files and git status. videos directories are ignored, but audio/raw directories are intended to be committed.

Script Behavior

• The default kind for --year-dir 2026-spring is lectures.
• The default input for --year-dir 2026-spring --kind lectures is 2026-spring/lectures/videos.
• The default output is 2026-spring/lectures/audio/raw.
• For nested videos, preserve the relative path under the selected content folder's audio/raw.
• Default output format is .m4a with AAC audio suitable for transcription workflows.
• Existing outputs are skipped unless --overwrite is passed.
• ffmpeg must be installed and available on PATH; the script reports a clear error if it is missing.
• Use --legacy-transcribations only when intentionally writing to the old transcribations layout inside the selected content folder.

Examples

Extract one file:

uv run python "$SKILL_DIR/scripts/extract_audio.py" \
  --repo "$REPO_DIR" \
  --kind lectures \
  2026-spring/lectures/videos/lecture_01.mp4

Extract seminar videos:

uv run python "$SKILL_DIR/scripts/extract_audio.py" \
  --repo "$REPO_DIR" \
  --year-dir 2026-spring \
  --kind seminars

Extract every lecture video in every year directory that has a lecture videos folder:

uv run python "$SKILL_DIR/scripts/extract_audio.py" \
  --repo "$REPO_DIR" \
  --kind lectures \
  --all-years