LinkedIn Video Editor for Claude Code: AI editing skill for talking-head videos
/video-edit-linkedin is a Claude Code skill that edits LinkedIn videos for you.
Drop in a raw vertical talking-head video (phone, webcam or green screen, any frame rate). You get back a finished 9:16
LinkedIn video with:
- a hook on the first frame;
- karaoke captions in your exact words;
- motion graphics and real photos;
- sound design;
- a LinkedIn Connect card at the end.
Every element is kept inside the LinkedIn safe zone, so nothing important hides under the app's UI.
You never open a video editor. Tell Claude which clip to edit, review the phone copy, and give notes in plain words.
Each round is one re-render.
It works in English and Hebrew, including right-to-left (RTL) Hebrew captions.
Quick start: install by pasting one message
No terminal needed.
Step 1: Open Claude Code
Open Claude Code: the Code tab in the Claude desktop app, or type claude in your terminal. The skill renders video
on your computer, so it needs Claude Code; a regular chat can't run it.
Step 2: Paste this message
Copy-paste this into the chat and hit Enter:
Install this skill for me: https://github.com/jacobggman/claude-code-linkedin-video-editor (into ~/.claude/skills/video-edit-linkedin)
Claude downloads the skill into your skills folder and tells you when it's ready. Takes a few seconds.
Step 3: Edit your first video
Type:
/video-edit-linkedin edit ~/Downloads/my-video.mp4
Hit Enter. On your first run, the skill does the following:
- asks your name, a profile photo for the LinkedIn card and your language (English or Hebrew);
- checks your computer;
- sets up everything it needs (a few minutes, once).
Just follow the prompts. Then watch the phone copy it sends you and say what to change.
Contents
What it does
| Feature | What you get |
|---|
| Hook from frame 0 | A bold headline that is readable on the very first frame, which is also the preview frame in the feed. Motion and sound arrive in the first second. |
| Karaoke captions | Word-by-word animated subtitles: the spoken word lights up and keywords keep their own colour. With your script, the captions show exactly what you said, not what speech-to-text heard. |
| Hebrew and RTL | Hebrew speech-to-text with ivrit.ai's Whisper model, right-to-left caption layout, Hebrew fonts (Heebo, Secular One, Suez One), and correct mixed Hebrew/English/number ordering. |
| Green-screen removal | Detects a green screen and keys it out in minutes (ffmpeg chroma key with despill). You get a clean cut-out of the speaker over a designed background for each section. With no green screen, AI background removal is used instead. |
| LinkedIn safe zone | Every card, word and face is checked against LinkedIn's overlays: the feed header, the post text and the action bar on either side. The build warns about anything that crosses them, and the final render is checked too. |
| Perfect lip sync, any frame rate | Frame-locked assembly converts 25 fps (or any rate) phone footage to 30 fps without drift. A cross-correlation sync check must pass before rendering. |
| Your take stays whole | Natural pauses are kept by default, so you still sound like you. Pause cutting is available on request. |
| Motion graphics | Animated cards, counters, dot grids, toggles, terminal windows, photo cards and words behind your head, rendered from HTML with HyperFrames and GSAP. |
| Head tracking | Reads where your head is through the take, so side-by-side layouts never put your face under the like/comment bar, and cards never cover your eyes, nose or mouth. |
| Sound design | 19 subtle, level-matched sound effects (whooshes, pops, clicks, risers), never louder than your voice. The final mix is mastered to -16 LUFS. |
| LinkedIn Connect ending | You leave the frame and a LinkedIn card with your real photo and name plays alone: "Let's connect on LinkedIn!", then Connect turns to Pending. No frozen last frame. |
Before and after
One continuous green-screen take, 25 fps, straight from the phone. The same moment in the finished video:
The LinkedIn video safe zone (9:16)
LinkedIn shows a 1080×1920 vertical video in two ways:
- In the feed: cropped to about 2:3, with your name and headline laid over the top and the post text and controls over the bottom.
- Full screen: an action bar (like, comment, repost, send) runs down one side, on the right in the English app and on the left in the Hebrew app.
These are the red lines this skill keeps important content out of. They were measured on screenshots of the LinkedIn app.
| Zone (1080×1920) | Pixels | What covers it |
|---|
| Top | y 0 to 340 | Name, headline, Follow; the feed crops the frame to about 2:3 here |
| Bottom | y 1620 to 1920 | Post text, "...more", sound and caption buttons |
| Action bar | x 0 to 160 and x 920 to 1080, from y 1100 | Like / comment / repost / send (left in RTL UIs, right in LTR UIs) |
| Side margins | 40 px | Rounded corners and crop |
Your face counts as important content: the skill places the speaker so the eyes-to-chin area stays clear.
build.py --safe draws these zones in red on preview snapshots, and scripts/safe_check.py paints them on frames
pulled from the final render.
This is the beginning of the README. Read the rest at the original. Read the full README ↗
