Lectures #

Local archive of lecture video/audio and machine-readable transcripts for AC-related talks.

Conventions #

Each lecture stores files under a single slug prefixed with the YouTube video ID (when available) so the origin is obvious:

<video-id>-<slug>.mp4   // highest-quality MP4 (gitignored, local-only)
<video-id>-<slug>.mp3   // audio-only MP3 for offline/quick playback (gitignored)
<video-id>-<slug>.vtt   // raw YouTube captions (VTT)
<video-id>-<slug>.md    // clean prose transcript with frontmatter

The large binaries (.mp4, .webm, .mkv, .mp3, .m4a, .wav) are gitignored; regenerate them locally with yt-dlp using the source_url in the markdown frontmatter.

For each lecture that should appear on papers.aesthetic.computer, a corresponding cleaned .txt lives under system/public/assets/papers/readings/text/ and is linked from system/public/papers.aesthetic.computer/platter.html — this is the same pattern used for other long readings (e.g. the Goodiepal Equation documentary transcript).

Regenerating a lecture from scratch #

cd papers/lectures
yt-dlp --write-auto-subs --write-subs --sub-langs 'en.*' --skip-download \
  --output "%(id)s-%(title)s.%(ext)s" <youtube-url>
yt-dlp -f 'bv[ext=mp4]+ba[ext=m4a]/b[ext=mp4]/b' \
  --output "%(id)s-%(title)s.%(ext)s" <youtube-url>
yt-dlp -x --audio-format mp3 \
  --output "%(id)s-%(title)s.%(ext)s" <youtube-url>

Then slug-ify filenames and convert the VTT to prose (dedupe rolling captions, strip inline <c> timing tags, reflow into paragraphs).