AI Video Effect · Short form
YouTube Clips
Turns one long horizontal video (podcast, interview, YouTube episode) into ranked vertical clips. Each clip has a context card on top, the speaker below, and a bold caption on the seam. You give Claude Code the recording; it plans the edit, renders it on your computer and exports one finished MP4 per clip.
# YouTube Clips — video style
Turns one long horizontal video (podcast, interview, YouTube episode) into ranked vertical clips. Each clip has a context card on top, the speaker below, and a bold caption on the seam.another 3,762 wordsEvery paid plan opens all 18 styles.
How it works.
- 01
Add your video
Put
episode.mp4in a folder and open that folder in Claude Code. - 02
Find the best 3 clips in episode.mp4 and edit them in the YouTube Clips style.Type the prompt
Paste it into Claude Code and press Enter. Claude writes a short plan first.
- 03
Get your edit
Claude edits it, checks it frame by frame and saves the finished MP4 in the Export folder.
First time? Unzip the skill into .claude/skills/ and type “Set up the YouTube Clips skill.” once.
Questions
What do I need to use YouTube Clips?
Claude Code (the desktop app's Code tab or the terminal), a Mac or Linux computer with Python 3.9+ and ffmpeg, and about 3 GB of free disk space for the speech and vision models, which download once. Then: one long horizontal video, 5 minutes to 3 hours: a podcast, interview or YouTube episode.
What does YouTube Clips export?
1080×1920 (9:16), 30 fps, H.264 + AAC, ready for Reels, TikTok and Shorts. One file per clip, plus CLIPS.md with the ranking.
Can I change the colours and fonts?
Yes. They live in the skill's brand settings. Tell Claude your colours and font, for example "Use my brand: #0E1A2B background, #F4EFE6 text, Gilroy font", and it updates them. The default fonts are free (SIL Open Font License).
Is my video uploaded anywhere?
No. Everything runs on your computer, and your video and voice are never uploaded. The only downloads happen during setup: Python packages, two MediaPipe models from Google, the Whisper speech model and the free fonts.
What happens when I correct something?
Claude adds your correction to the skill as a rule, so the same thing does not happen in your next edit.
What should I prepare?
One long horizontal video, 5 minutes to 3 hours: a podcast, interview or YouTube episode. Optional: the episode title and guest names.
What does YouTube Clips not do?
The source has to be one horizontal video between 5 minutes and 3 hours long. It runs on macOS or Linux with Python 3.9+ and ffmpeg, not on Windows. Sound effects are not included. Add your own licensed samples, or it uses clean built-in sounds.
More like this.
- Business ReelsFull screen, split screen and full-screen animation, switching on every cut.
- Explainer PanelYou in a panel at the bottom, animated explanations on top.
- B-roll and Stacked TextTight cut, B-roll, captions with stacked key phrases.
- Fitted CaptionsOne big accent word with the rest fitted around it.
- Cut-out BackdropYou cut out over full-screen images that change with every idea.
- Cut-out GridYou in the centre of four images that change together.
- Depth UIYou full frame, real interfaces and screens behind your head.