A skill that makes Claude board a video the way a senior director does, one stage at a time, with you signing off each one. It sets the specs for every placement, times your script into beats and counts the voice-over against the clock. Then it writes every frame: number, timecode, shot size, angle, camera move, action, voice-over, on-screen text and sound. It works out the 9:16, 4:5 and 16:9 versions and the 15 and 6 second cutdowns, writes image prompts for each frame if you want pictures, and checks the whole board before it goes to a crew.
Inside are four reference files. One holds the lengths, shapes, safe zones in pixels and crop maths for Meta, TikTok and YouTube, the FTC's disclosure guidance, caption and flash limits, and the limits of Midjourney, GPT Image, Nano Banana and Veo, each with its source and the date it was checked. The others hold the timing rules (2.5 words a second for voice-over, a hold-time formula for on-screen text), the camera grammar, the prompt formula with a consistency kit, and a full board for a fictional rain jacket brand. Claude never invents a claim, price, review or testimonial, and it marks every frame that must be real footage.
Claude cannot film or make images in the chat. It writes prompts for your own tools, and generated frames are for planning only. Switch on code execution so it can count words and seconds exactly and make the PDF board, the shot sheet, caption files and a client deck; without it you get tables and text. The platform figures were checked on 17 October 2026 and change without notice, so look at your ad manager's preview before a big spend. Bring the script, where it will run, product photos and the brand rules, and start at whichever stage you have reached.
Claude skill · version 1.0
Video storyboard
Checking your access.
In the download · 6 files
- SKILL.mdThe method Claude follows: specs and master ratio, beats and timing, the frame-by-frame board, versions and cutdowns, frame prompts, and the final check and hand-off.
- README.mdHow to install it in claude.ai or Claude Code, how to start, and the licence.
- references/formats-and-specs.mdLengths, shapes, safe zones in pixels and crop maths for Meta, TikTok and YouTube, plus loudness, captions, FTC disclosure rules and image and video tool limits, each with its source, checked on 17 October 2026.
- references/timing-and-structure.mdVoice-over words per second, the hold-time formula for on-screen text, shot lengths, structures for 6 to 90 seconds, shot sizes, lens feel, camera moves, continuity, transitions, sound layers and board notation.
- references/frame-prompts.mdSketch or photoreal frames, the consistency kit, the frame prompt formula, tool settings per ratio, the animatic, and the layouts for the PDF board, shot sheet, caption files and client deck.
- references/worked-example.mdA full 30-second board for a fictional rain jacket brand, with its 15 and 6 second cutdowns, three ratios, frame prompts and the final check.
Install it and begin
01Install it in Claude CodeTerminal mkdir -p ~/.claude/skillsunzip ~/Downloads/video-storyboard.zip -d ~/.claude/skills/# Start it with /video-storyboard. If this is your first skill, type /reload-skills once.On claude.ai there is nothing to unzip: upload the zip as it is. Every step is on Install and use a Claude skill.
02Your first message to ClaudeText Use the video storyboard skill. Here is my [SCRIPT OR IDEA]: [PASTE IT HERE]. It is a [LENGTH IN SECONDS] [AD, REEL OR BRAND FILM] for [WHERE IT WILL RUN]. We will make it with [LIVE ACTION, AN AI VIDEO TOOL OR BOTH]. Start with the specs and the master ratio.Attach product photos, your logo and any brand guidelines. If you already have a board in one shape, attach it and ask for the other versions instead.
Worked example
How a director turns a script into a board
- 01Write the one thing the film must land, then list every placement with its length, shape and pixel size.The master ratio follows the money. If most of the spend is Reels and TikTok, compose in 9:16 and treat 16:9 as the version.
- 02Count the voice-over against the clock at 2.5 words a second, leaving 1.5 seconds clear.A 30-second spot holds about 71 words at most. Cut words before the shoot; never plan to speed up the read.
- 03Board every frame with its timecode, shot size, angle, move, action, voice-over, on-screen text and sound.A blank field is a decision someone makes on the day. The last frame's out time must equal the length, to the tenth of a second.
- 04Hold each line of on-screen text for 1 second plus 1 second for every 3 words, never under 1.5 seconds.Viewers did not choose to read it. Five words need 3 seconds, and a legal line gets one more.
- 05Mark every frame that shows what the product does as real footage only.A generated or sped-up demonstration presented as real is a deceptive mock-up under US advertising law.
- 06Work out each version frame by frame: keep, reframe, its own shot, or drop.A 9:16 crop from a 16:9 frame keeps 31.6% of the width. People shots rarely fit, so a second camera set horizontal often costs less than a second day.
- 07Write frame prompts from one consistency kit, with no words in the picture.The same style, character and product lines in every prompt keep frames matching. Supers and logos go on in layout, because image tools misspell them.
What to do it with
- ClaudeFree tier, paid above it
A language model that writes, edits, reads long documents and runs tools, including agents that work inside a codebase.
Bad at It is agreeable by default, so it will improve a bad idea instead of telling you it is bad. Ask it to argue against you and the quality of the answer changes.
- Claude CodeSubscription
An agent that reads a whole repository, edits files, runs the tests and explains what it changed.
Bad at It is fast enough to make a large mistake quickly. Work on a branch, read the diff, and never let it push. Watch the starting mode: on Pro, Max and Team plans it opens in auto mode, which Anthropic's own list says may read a .env file and send the key in it to the API that key belongs to. Switch it off.
- MidjourneySubscription
Generates images from a written description, with the strongest default aesthetic of anything in the category.
Bad at That default aesthetic is the problem as often as it is the draw. It has a house look, it pulls everything towards it, and getting a plain honest photograph of a real product out of it is harder than it should be. Text inside an image is unreliable.
- GPT ImagePay per use
Generates and edits images with unusually good instruction-following, and can hold text inside a picture better than most.
Bad at Faces drift between generations unless you feed a reference every time. Skin can come out plasticky at default settings, and a heavy fidelity setting costs many times more and takes many times longer.
- VeoPay per use
Text or image to video, with sound, at a length and coherence nothing had two years ago.
Bad at Cost per second is high enough to change how you storyboard. Long shots still drift, and hands and logos wander. Through the Gemini API on 25 September 2026 it refused a last frame on every tier, though the docs describe one.
Why it matters
Ask a chat for a storyboard and you get a list of nice shots with no clock. The voice-over runs eight seconds long, the on-screen text is gone before anyone can read it, the 9:16 version is cropped from a wide frame that never fit, and the product demonstration gets generated when it had to be filmed. This skill makes Claude time every frame and every word, plan each shape and cut before the shoot, and mark what must be real.
- Type
- technique
- Period
- ongoing