YouTube to Markdown: Clean, Token-Efficient Transcripts for Obsidian & Claude

By · Updated 2026-10-19

Paste any public YouTube video or finished livestream replay URL to export clean, timecoded Markdown:

120 free Credits on signup · Captions-first (~20s) + automatic audio ASR fallback when CC is missing

Why copying from YouTube's transcript panel breaks your notes and AI prompts

When building a personal knowledge base in Obsidian or feeding a 90-minute technical talk into Claude, the formatting of your source text matters as much as the words themselves. Standard methods produce messy artifacts:

  • Raw YouTube "Show transcript" copy-paste — YouTube splits speech into fragmented 4-to-6-word visual caption lines, places every timestamp on its own separate line, and often lacks sentence punctuation on auto-generated tracks.
  • Raw .srt / .vtt subtitle files (+50% token bloat) — Subtitle files repeat sequential line numbers and millisecond arrow intervals (00:14:02,120 --> 00:14:05,480) every few seconds. As measured in our podcast token benchmark, a 1-hour talk that takes ~11,500 tokens in clean Markdown balloons to ~17,800 tokens in SRT — wasting over 6,000 context tokens on formatting noise.

Side-by-side: Raw YouTube copy vs. SRT vs. VideoScript Markdown

FormatActual Output Structure1-Hr Token Cost
YouTube Panel Copy0:14\nso when we look at gradient\n0:17\ndescent in non-convex landscapes~14,200 tokens (broken lines)
Raw .SRT Subtitle4\n00:00:14,100 --> 00:00:17,450\nso when we look at gradient descent~17,800 tokens (+50% bloat)
VideoScript Markdown[00:14] So when we look at gradient descent in non-convex landscapes...~11,500 tokens (optimal)

Works when captions are missing — and on finished livestream replays

Browser extensions and free "YouTube to Markdown" bookmarklets only scrape YouTube's existing caption API. The moment a creator disables captions, or an academic lecture has no CC track, those tools return an empty error.

VideoScript solves this with a dual-path architecture:

  1. Captions-first fast path (~20 seconds) — If the video carries a valid caption track, VideoScript retrieves it, strips cue noise (such as >> markers), and formats clean timestamped segments in roughly 20 seconds.
  2. Automatic audio ASR fallback (~50x real time) — When captions are missing or disabled, VideoScript automatically extracts the audio stream and runs high-accuracy speech-to-text transcription — no 2 GB video downloads required on your machine.
  3. Finished YouTube livestream replays supported — Paste any completed youtube.com/live/... or archived webinar replay link to get the full transcript. (Active ongoing livestreams are cleanly rejected before any Credits are spent.)

Workflows for Obsidian, Notion, and Claude

  • Obsidian second brain — Export as .md and drop the file straight into your vault's Sources/ folder. Because every segment starts with [MM:SS], you can wrap key passages in > [!quote] callouts and link concepts with [[Wikilinks]] while keeping exact timestamp provenance. See our Obsidian & Notion lecture notes guide.
  • Claude & ChatGPT analysis — Copy the Markdown transcript into Claude (200k context window) or ChatGPT and ask for grounded takeaways with mandatory [MM:SS] citations so you can verify every claim against the video. See how to feed YouTube transcripts into Claude & NotebookLM.
  • Plain text & subtitle exports included — Need text without markdown headers or need edit-ready captions? Every job also exports as Plain Text (.txt), .srt, and .vtt.

Convert any public YouTube video to Markdown

Get 120 free Credits on signup (covers 120 minutes of video) with no credit card required. One-time Credit packs never expire.

Start Converting Free →

Frequently asked questions

  • How do I get a YouTube transcript formatted for Claude or Obsidian?
    Paste the public YouTube URL into the box above (or into the VideoScript workspace). Once processed, click Export → Markdown (.md) or copy the formatted text directly into your Obsidian vault or Claude prompt.
  • Does this work on YouTube videos that have no CC subtitles?
    Yes. When a video has no captions or the creator turned off "Show transcript", VideoScript automatically falls back to transcribing the audio track at roughly 50x real-time speed. Read more in our YouTube transcript guide.
  • Can I convert a finished 3-hour YouTube livestream replay to Markdown?
    Yes. Completed livestream replays (youtube.com/live/... or standard watch URLs) are fully supported. Only currently broadcasting live streams, private videos, and members-only / age-restricted videos are excluded.
  • Does VideoScript generate an AI summary instead of the full transcript?
    No. VideoScript always delivers the complete, verbatim, timecoded transcript — never a truncated summary. Read why in Transcript vs. summary: which do you need?