@profullstack/media2markdown-core

The pipeline behind media2markdown: video, audio and PDF into a chaptered Markdown bundle, using open-source tools only.

import { convert } from "@profullstack/media2markdown-core";

const result = await convert({
  input: "https://www.youtube.com/watch?v=...",
  outDir: "./lecture",
  zip: true,
  onProgress: ({ stage, message }) => console.log(stage, message),
});

Chapter boundaries

Boundaries come from the best evidence available, never from a guess:

1. Container chapters — ffprobe -show_chapters (mkv, m4b, mp4)
2. Publisher chapters — yt-dlp info.json
3. Document outline — headings from the PDF parser
4. Long silences — ffmpeg silencedetect
5. Fixed windows — last resort

result.chapterSource records which tier was used.

Also exported

doctor, installFfmpeg, installWhisper, probe, extractAudio, extractKeyframes, detectSilence, transcribe, ocrKeyframes, pdfToMarkdown, planChapters, writeBundle, zipBundle.

MIT © Profullstack, Inc.