FREE SKILL
Steal This Skill: Six Reel Formats From One Chat
Type a format name, paste the words you say in a 3 to 5 second clip, and Claude hands back the whole edit: where everything sits on the frame, the word cards with their timings, the on-screen graphic as a file you open in a browser, and a cut sheet for CapCut, Premiere or DaVinci Resolve.
Six formats. One skill. Nothing to install.
The pack is at the bottom of this page. Everything it knows is also on this page, so you can build the first one before you download anything.
The Problem: Every Reel Starts From Zero
You film a clean 4 second line. Then you open the editor. The next hour goes on decisions: where the title sits, how big the captions are, when the graphic appears. Does the card cover your mouth?
Next week you do it again, from zero, for a reel that looks almost the same.
The reels that work right now use the same six formats. A page behind the head. A tier list on the chest. Price tags on B-roll. Phones fanned around the face. Title boxes over a call clip. A diagram on top, you below. The formats are not the problem. Nobody writes them down, so every edit is a rebuild.
This pack writes them down. Pixel positions, timings, caption rules, and the graphic itself. Claude reads the sheet and fills it with your clip.
What Is a Reel Format?
A reel format is a 3 to 5 second unit. You on camera saying one line, one graphic that appears on cue, and word cards that follow your speech. Every unit has the same beat pattern. The graphic enters at a tenth of a second. A second element arrives at around 0.7 seconds. Something changes at about 60 percent of the clip length. Then it holds while the sentence finishes. Chain five or six units and you have a reel.
The Six Formats
All positions are on a 1080 by 1920 frame, measured from the top left. Keep anything that must be read between 250 and 1600 pixels from the top, because the apps draw their own buttons over the rest.
Depth UI. A screenshot of a web page floats behind your head in a rounded window (1000 by 625 pixels, 150 from the top). You are cut out of your background and placed in front of it. A card with a name, a link and a button sits on your chest at 1235 from the top. At 2.2 seconds the button flips from JOIN to JOINED. Captions at 1118. Use it when the line mentions a page, a product, a community.
Ranking. A two-line title at the top on a dark gradient, the second line in your accent color. An S to F tier table on your chest from 930 down, each row 62 pixels with a colored label. One item pops up large in the centre at 0.7 seconds, then shrinks to 17 percent and drops into its row at about 2 seconds. Captions at 838, between the title and the table. Use it when the line ranks or compares.
Personal B-roll. Your own footage fills the top 1150 pixels. Up to four price tags pop onto it, zigzagging down the frame at 0.45, 1.05, 1.6 and 2.2 seconds. You talk in the bottom 770 pixels. Captions on the seam at 1080. Use it for money, costs, numbers, or whenever you have footage of the thing.
Business. The tightest framing, scaled 15 percent. A two-line title at 165 from the top, the second line in a white box. Phone screenshots in seven slots around your head, popping in one by one from 0.35 seconds and floating 9 pixels up and down after. Captions are single big words at 94 pixels. Use it for a promise or a claim with examples to show.
Coaching call. Two white title boxes stacked at 150 from the top. You in the middle, usually with a mic in frame. Uppercase display captions at 1010 with one italic accent word per unit, never two. Use it for clips cut from a recorded call or lesson.
YouTube clips. A dark card on the top 900 pixels with a diagram on it. The default diagram is nine avatar circles in a grid, the centre one lit in your accent color. The circles pop in one by one, and the lit one pulses at 1.2 seconds on the word it stands for. You in the bottom half. Captions on the card at 790. Use it for people, teams, steps.
How It Works in the Chat
Install the pack. The start file in the folder shows both ways. Drop it into Claude Code's skills folder, or upload the zip under Settings and Capabilities in the Claude app. Then say something like this:
Use the AI editing pack. Format: ranking. My clip says: "Range, fields you'd never think to use." It is 3.5 seconds. The ranking is best channels to sign clients: Instagram S, YouTube A, email B, TikTok drops into S.
Claude asks for anything missing, in one message. Then it returns the cut sheet, the word cards with their start times, and the graphic as a complete HTML file. In Claude Code it saves the file next to your clip. In the app it gives you the code to save as graphic.html.
Open the file in Chrome with ?bg=green added to the address. It animates once on load. Record it at 1080 by 1920, or take a screenshot if nothing moves after the entry. In your editor, key out the green, stack it over your clip, and add the word cards as text layers from the list.
The first unit is slow because everything is new. By the third you are mostly pasting.
Copy This Prompt
If you want to try one format before installing anything, paste this into any Claude chat. It carries the ranking format's rules inside it.
Copy this.
You are my reel editor. I will give you the words I say in a 3 to 5 second vertical clip and what the graphic should show. You return the edit for the RANKING format, on a 1080 by 1920 frame, pixels from the top left. The format: - A two-line title at top 175, centred, 78 px, uppercase, heavy condensed italic. Line one white, line two in [MY ACCENT COLOR, e.g. #7C3AED]. A dark gradient over the top 560 px so it reads. - A tier table at left 110, top 930, width 860: rows S A B C D E F, each 62 px high, a 72 px colored label (S #FF7F7F, A #FFBF7F, B #FFDF7F, C #7FFF7F, D #7FBFFF, E #BF7FFF, F #FF7FFF), dark cells with 110 by 50 chips for the items already placed. - A pop card 520 by 300 at left 280, top 920: the item's name at 68 px uppercase in the accent color. - Word cards: white, 900 weight, 58 px, centred at top 838, one to three words each, one card at a time, one accent word at most. The beats: gradient at 0.05 s, title at 0.10, table at 0.25, pop card at 0.70 (from half size, slight overshoot), the card shrinks to 17 % and slides into its row at 60 % of the clip length. Hold after that. My clip: [THE EXACT WORDS I SAY] Clip length: [SECONDS] Title: [LINE ONE] / [LINE TWO] Items already placed: [ITEM in ROW, ITEM in ROW] The item that drops: [ITEM into ROW] Return, in this order: 1. A layers table (clip, graphic, captions) with positions. 2. A beats table with the times for my clip length. 3. The word cards: words, start, end, accent (group the words the way they are spoken, breaking on commas and stressed words; if I give you auto-caption timings, use them; otherwise spread the cards by syllables). 4. One complete HTML file, 1080 by 1920, that draws the title, the table and the pop card with CSS animations at the beats above, a green background when the address ends in ?bg=green, no external scripts. 5. The assembly steps for [CAPCUT / PREMIERE / RESOLVE]. End with "Done."
Swap the format block for any of the other five using the numbers in the section above. The pack has all six written out, plus the graphic files already built.
What You Get Back
A cut sheet: layers, pixel positions, and the second each thing appears, computed for your clip length.
The word cards: your words grouped the way you say them, with start and end times and the one accent word marked.
The graphic as a real HTML file that animates on load, with a green background option for keying.
The assembly steps for your editor, down to the chroma key setting.
A reusable unit. Swap the words and the graphic content and the same sheet edits the next clip, and the one after that.
Honest Limits
Claude does not render video in the chat. It writes the sheet and the graphic; you stack them in your editor. If you use Claude Code with ffmpeg and a headless browser installed, it can render the graphic to a video file for you. That is optional. The browser recording gives the same result.
The caption timings are estimates unless you paste the auto-caption times from your editor. Check them against your mouth once and nudge the late ones by a tenth of a second.
The depth UI format needs your editor's background removal to cut you out. On a plain wall with front light it is clean. Against a busy room it leaves a halo, and no sheet fixes that.
The graphic files ask for two free fonts from Google Fonts and fall back to system fonts offline. The look is close either way.
If you want Claude to make the whole video rather than the edit, the motion design skill is the other build. This one is for clips you filmed yourself.
Do This First
Film one 4 second line against a plain wall. Run the ranking format on it with the prompt above or the pack. Assemble it in whatever editor you already have.
Twenty minutes. If you do not like the result, you lost twenty minutes and kept a sheet that edits every reel after it.
Done.