Working with YouTube videos
VidNotes plays YouTube videos via embed. Auto-transcribing them is hit-or-miss because of macOS sandbox restrictions and YouTube's own anti-scraping measures — so VidNotes gives you three paths to a transcript, all of which unlock the full AI feature set (Brief, Chat, Synthesis, Search).
- App Sandbox — VidNotes ships through the Mac App Store, which forbids spawning external tools like yt-dlp. We can't download the audio for you.
- YouTube PoToken — late-2025 change that blocks most third-party caption scrapers from reaching auto-generated tracks. Public manual captions still work; auto-captions usually don't.
Paste the YouTube URL, click Transcribe. If the video has manually-uploaded public captions (educational channels like MIT OpenCourseWare, news organizations, official corporate uploads), the transcript fills in immediately.
If captions exist, you're done — Brief, Chat, Synthesis, and Search all light up.
The empty-state panel will show three cards. Pick the one that fits your situation:
Save the YouTube video as MP4 using yt-dlp on your terminal, 4K Video Downloader, a browser extension, or any tool you prefer. Drop the file into VidNotes — WhisperKit transcribes it on-device. No cloud calls, full quality.
Trade-off: requires a one-time setup of a download tool the first time you use it.
YouTube's own "…" menu has a Show transcript option that opens a panel of timestamped lines. Select all, copy, paste into VidNotes. Done in 10 seconds.
Trade-off: you're using YouTube's auto-captions, which can mishear technical terms — but for most lectures, perfectly good.
VidNotes integrates with youtube-transcript.io. Sign up for a free account (50 transcripts/month), generate an API token, paste it into Settings → AI Providers → youtube-transcript.io. After that, the empty-state card becomes a one-click "Fetch now" for any YouTube video without public captions.
Trade-off: third-party dependency, free tier capped at 50/month. Heavy users pay their tier upgrade. Privacy: only the YouTube video ID leaves your Mac, never the audio.
Even without a transcript, you can take timestamped notes. As of v1.1, VidNotes can generate a Brief, run Chat, and answer questions from your notes alone — no transcript required. So if you'd rather watch and write, that's a valid path.
Click Brief → Generate. With on-device Local AI (Apple MLX, no API key needed) or any cloud provider you've configured, you get the four-layer research brief just like for local files.
brew install yt-dlpThen download any YouTube video as MP4 + audio:
yt-dlp -f "best[ext=mp4]" "https://youtube.com/watch?v=YOUR_ID"Drop the resulting file into VidNotes. Don't have Homebrew? Install it first — takes 2 minutes.