video-shotcraft: a Remotion shot library that turns Claude Code and Codex into a motion-design studio
AI video skill for Claude Code & Codex — cinematic product videos with Remotion: 152 shot recipe cards, 209 motion previews, a production-ready template
At a glance
- What is it?
- video-shotcraft is an agent skill for Claude Code and Codex that storyboards, animates and sound-designs product promos in Remotion. It ships a 36.2-second Ink Press template, a browser Motion Workbench and a CapCut CN export path, but its documentation assumes you already have an agent and a Node toolchain.
- Who is it for?
- Adopt video-shotcraft if you already run Claude Code or Codex, your product has a live page worth capturing, and you want a Remotion project you can hand to a designer afterwards. Skip it if you need narration-driven timing (that is video-talkcraft), if you cannot run a Node toolchain, or if you expect the skill to invent a brand identity from nothing.
- Can I use it commercially?
- Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 1 day ago.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 29, 2026, and from our analysis. They are not legal advice.
Editorial analysis
The gap video-shotcraft fills between a script and a finished promo
Most teams that need a 30-second product video face the same split. A motion designer can build it, but the queue is weeks long and every revision costs another pass. An editor can cut something in CapCut, but the result is screenshots sliding across a timeline. video-shotcraft targets the space between: an AI agent that already knows how to write code, given a library of named shot recipes and a Remotion project to put them in.
The intended user is an engineer or technical founder who already has Claude Code or Codex open. The README's quick start is addressed to that person directly: hand the repository link to your agent and say install this skill. From there the requests are conversational. The README gives examples like using the deck-deal-flyin and row-embed shot cards to present a feature, or designing a close-up inspired by spotlight-hero-card. The output is a Remotion composition, not a rendered file you cannot touch again.
The scope is deliberately narrow. This is a product promo, marketing, launch or demo video, in a paper-ink-amber visual register with 2.5D camera moves over real page captures. It is not a talking-head explainer and not a general video editor. The README points narration work at a sibling project, video-talkcraft, which handles word-level timestamp alignment and a separate 78-card library.
Shot recipe cards, Remotion components and the Ink Press template
The mechanism is a three-layer stack, and the README describes each layer in enough detail to reason about it.
The top layer is the shot recipe card. The README counts 152 cards and 209 motion previews, each card carrying a name, a description and a preview you can browse in the online Gallery. Cards are the vocabulary you speak to the agent: deck-deal-flyin, row-embed, spotlight-hero-card. The README says the 48 cards added in 2026-08 were distilled from 209 candidate motions through eight rounds of frame-by-frame review against reference footage, which tells you the curation was manual rather than scraped.
The middle layer is the implementation. Each recipe maps to a native Remotion component at demos/<category>/<name>/<Component>.tsx. The README states these are deterministic and driven by a normalized progress value t, with a wiring snippet documented in demos/README.md. That determinism matters for a reason the README does not spell out: Remotion renders frames independently, so any component that reads wall-clock time or random state will produce a different film on every render. Pinning everything to t is what makes a re-render reproducible.
The bottom layer is the template. Ink Press is a complete promo: 36.2 seconds, 1920x1080, 30fps, 10 shots, with title cards, transitions and what the README calls a fully pinned cinematic SFX pass. The README notes the 38-second Gallery intro was itself produced with the skill, so the project has at least one end-to-end artifact it can point to.
One design choice is worth flagging. The README describes the cards as de-branded, with neutral placeholder copy and a single swappable ACCENT color variable. That is a deliberate constraint: you get structure and motion, and you supply the identity. If you were hoping the skill would propose a palette and typography system, that is not what the cards do.
Installing video-shotcraft and running a first promo
There are three install paths in the README, and they are not equivalent. The first is conversational: in Claude Code, Codex or a similar agent, paste the repository link and ask the agent to install it. The README states the agent will clone the repository and link it into your skills directory.
The second uses the skills CLI, which is the shortest path if you would rather not delegate:
npx skills add Vincentwei1021/video-shotcraftRunning that should place the skill where your agent looks for skills. The README does not document what the command prints on success, so check your skills directory rather than trusting the exit code alone.
The third is manual, and it is the one that tells you most about how the skill is wired. The README clones the repository and creates a symlink into the agent's skills directory:
git clone https://github.com/Vincentwei1021/video-shotcraft.git
cd video-shotcraft
ln -s "$(pwd)" ~/.claude/skills/video-shotcraft # Claude Code
# or
ln -s "$(pwd)" ~/.codex/skills/video-shotcraft # CodexBecause it is a symlink, editing files inside the cloned repository changes what the agent sees on the next request. That is convenient for tweaking a component and awkward if you later move or delete the clone, since the symlink breaks silently.
With the skill linked, the first real use is a request, not a command. The README suggests telling the agent to make a promo for your product with the Ink Press template. The README also states that if you do not name a shot card, the skill introduces the built-in template first and asks whether to use it, so you can start with a vague request and let the agent narrow it. You can also browse the Gallery beforehand and copy shot-card names into your prompt.
Once a film is delivered, the README describes a second stage: the skill opens a Motion Workbench in the browser.
node workbench/scripts/open.mjs <project>The workbench decomposes the film into shot, transition, caption and SFX tracks. The README says you can edit copy, font sizes and colours in a schema-driven inspector, move, trim or speed-ramp clips, drag in any of the 216 demo motions, and export with Remotion. It claims preview and render are frame-identical, described as pixel-parity verified. That claim is the project's own; it is the kind of thing worth re-checking on your footage before you rely on it.
Where video-shotcraft breaks down
The most concrete limitation is the one the README admits by pointing elsewhere. video-shotcraft is built for motion-led product films, and the sibling project video-talkcraft exists specifically because narration videos need a different timing model. If your video is a voiceover with visuals underneath, word-level alignment is the problem you have, and this skill does not solve it. The README states video-talkcraft aligns timestamps locally with a median of 20 to 40 milliseconds per character and ships its own 78 recipe cards. Choosing the wrong one of the two means fighting the tool.
The second constraint is environmental. The repository's package.json is not a runtime manifest for the video pipeline. It is named video-shotcraft-tests, marked private, and describes itself as unit tests for assets/lib/helpers pure functions, with vitest as the only dependency. The README does not document a Node version, a Remotion version, or a system dependency list for rendering. Remotion itself needs a headless browser to render, and nothing in the README states which one or how it is provisioned. Treat the render environment as your responsibility until you have confirmed otherwise.
The third is the export path. JianYing project export is documented as verified on JianYing Pro 11.2 for macOS. That is a single platform and a single application version. If you are on Windows, or on a different JianYing build, the README gives you no basis to assume the draft will import cleanly. The same caution applies to the Workbench's frame-parity claim: parity between preview and render is exactly the property that breaks first when a component starts depending on something outside t.
Finally, the recipe library is opinionated. The cards are de-branded and the style is paper-ink-amber with 2.5D page captures. If your product is a dark-mode developer tool or a mobile app with no web surface, the shot vocabulary will still work, but the visual register the template establishes may not be the one you want, and re-tuning it is work the README does not quantify.
video-shotcraft against a plain Remotion project
The honest alternative is Remotion itself. Remotion is the rendering engine video-shotcraft is built on, and it is a general framework: you write React components, declare compositions, and render them to video. Nothing about video-shotcraft is a fork of Remotion or a wrapper that hides it. The template is a Remotion project, and the shot components are Remotion components.
The difference is what you start with. A plain Remotion project gives you an empty timeline and a rendering model. You decide what a shot is, how transitions compose, how sound effects land on cuts, and what a good 2.5D camera move looks like. That is freedom, and it is also the entire cost. video-shotcraft front-loads those decisions into 152 named cards with previews you can watch before committing, plus a 36.2-second reference film that demonstrates how the pieces fit together.
The second alternative is a timeline editor. CapCut, JianYing and their peers give you direct manipulation and instant feedback, which no code-driven pipeline matches for a two-minute cut. video-shotcraft does not compete there. Its JianYing export is a handoff, not a replacement: the README describes the plate being cut per shot, captions rebuilt as native text tracks, and SFX and BGM landing on separate audio tracks, so you can retime, reorder and grade in the editor. The interesting property is that the draft is generated from an authored structure, so the tracks arrive organized rather than as one flattened render.
If your team has no motion designer and no appetite for writing React, a template-driven editor is still the faster route. video-shotcraft assumes you want the film to live in a repository, under version control, re-renderable when the product changes.
Licence, maintenance and what an upgrade actually costs
video-shotcraft is Apache-2.0. For a skill that generates video assets, that matters in two directions. Apache-2.0 includes an express patent grant and requires that you preserve the licence and attribution notices when you redistribute the work or derivatives. The shot components you adapt into your own Remotion project are derivatives in the ordinary sense, so keeping the notice is the safe reading. This is not legal advice; if you are shipping the output commercially at scale, have someone check how your use interacts with the licence and with Remotion's own licensing terms, which are separate and not covered by this repository's LICENSE file.
The repository is not archived, and the last push was on 2026-09-09, which is recent. The release history shows two asset drops: gallery-media on 2026-07-26 and showcase-media on 2026-09-01. Both are media, not code. That pattern is worth naming, because it tells you where the project's energy goes. The library grows by adding previews and cards; the engine underneath is Remotion and moves when Remotion moves.
Upgrade cost therefore splits. Pulling new shot cards is close to free: they are additive components you can ignore until you want them. Pulling a new Remotion major version is not, because the cards are React components compiled against it, and the README does not document a version pin or a migration path. The repository ships package-lock.json, which pins the test toolchain, but the README does not show a lockfile governing the video pipeline's Remotion dependency. Before you build a long-lived project on this, check which Remotion version the template expects and decide whether you are willing to track it.
Editorial conclusion
Adopt video-shotcraft if you already run Claude Code or Codex, your product has a live page worth capturing, and you want a Remotion project you can hand to a designer afterwards. Skip it if you need narration-driven timing (that is video-talkcraft), if you cannot run a Node toolchain, or if you expect the skill to invent a brand identity from nothing. Before committing, verify three things yourself: that `npx skills add Vincentwei1021/video-shotcraft` resolves in your skills directory, that the Ink Press template renders at 1920x1080 and 30fps on your machine, and that the Motion Workbench opens with `node workbench/scripts/open.mjs <project>` and previews the same frames you get on export.
Frequently asked questions
What is a shot in video-shotcraft?
In video-shotcraft a shot is a named recipe card, such as deck-deal-flyin or spotlight-hero-card, backed by a native Remotion component at demos/<category>/<name>/<Component>.tsx. The README states these components are deterministic and driven by a normalized progress value t, so each shot renders the same frames every time.
How do I install the video-shotcraft skill in Claude Code or Codex?
The README gives three routes: tell your agent to install the repository link, run npx skills add Vincentwei1021/video-shotcraft, or clone the repository and symlink it into ~/.claude/skills/video-shotcraft or ~/.codex/skills/video-shotcraft. The manual route uses a symlink, so moving the clone later breaks the link.
Does video-shotcraft work for narration videos with a voiceover?
No. The README directs narration work to a sibling project, video-talkcraft, which locks motion beats to word-level voiceover timestamps and ships its own 78 recipe cards. video-shotcraft is aimed at motion-led product promos, launch videos and demos.
Can I keep editing the video after video-shotcraft delivers it?
Yes, in two ways. The README states the skill opens a CapCut-style Motion Workbench with shot, transition, caption and SFX tracks, and that the film can also be exported as an editable JianYing draft with captions rebuilt as native text tracks. The JianYing export is documented as verified on JianYing Pro 11.2 for macOS.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/vincentwei1021-video-shotcraft)