Turn Claude Code Into a Motion Studio: How video-shotcraft Grows Product Promo Videos on Its Own

For most developers and independent teams, making a product video is a real hurdle. Outsourcing is expensive, and learning to edit yourself eats up time. Even after you finally cut something together, it still feels like it is missing something — the camera moves, the rhythm, that cinematic polish. More realistically, by the time the video is done, the product’s launch window has often already passed.

But a recent open-source project takes a very different approach: instead of making you learn video editing, it turns your AI coding agent (Claude Code, Codex) directly into a “motion design studio.” You give it product screenshots and copy, and it handles the storyboard, animation, and sound effects end to end, producing a cinematic product promo video for you.

It is called video-shotcraft.

Not Another “Text-to-Video” — But a Shot Recipe Library Plus a Pipeline

At first glance, many people mistake it for yet another AI video generator. In fact, its positioning is completely different. video-shotcraft is an agent skill, and under the hood it runs Remotion — the open-source framework that writes video with React components and renders on the server. In other words, what it produces is not mystifying generated footage, but controllable, reproducible finished videos that can go into your codebase.

Its core asset is 106 “shot recipe cards.” Each card reads like a motion-design spec: the shot’s purpose, its energy, a suggested duration, specific parameters, implementation notes, and even known pitfalls. Alongside them come 161 motion previews covering 162 styles, all housed in an online gallery that supports search, filtering, and one-click copying of card names. For example, deck-deal-flyin is a card that flies feature cards into the frame like dealing playing cards, while row-embed arranges several features in a row and lights them up one by one. You can even mix and match multiple cards to assemble a complete launch video. Behind each card is tuned TSX source code, with easing and timing parameters that actually run — not just illustrations.

A Six-Stage Pipeline: From Screenshot to Finished Video

Its real strength lies in methodology. The repository distills a six-stage production process: capture → visual direction → storyboard → sound design → beat sync → final QA. The pipeline also emphasizes deterministic rendering and independent subagent review, striving to make every output stable and reproducible rather than a roll of the dice each time.

Take “capture”: it ships with page-asset capture scripts that turn your real interface into video material. Take “sound design”: the sound assets come from Mixkit’s royalty-free library, with an emphasis on frame-accurate SFX — audio strictly aligned to the visuals. And “beat sync” has a dedicated methodology teaching the AI how to cut in time with the background music.

Experience that once lived only in the heads of seasoned motion designers has been broken down into executable rules written into SKILL.md and references, so the agent can follow along.

The Ready-to-Use Ink Press Template

If even picking cards feels like too much, the project bundles a complete, pre-validated template: Ink Press. It is 36.2 seconds long, 1920×1080, 30fps, with 10 shots, an ink-and-amber style, 2.5D camera motion, and frame-accurate sound effects. Components like transitions, subtitles, and number counters are all wired up in the template — what you change is the content, not the engineering. Swap in your product screenshots, copy, and brand colors, and you can reproduce a finished video.

This is especially friendly to indie developers and small marketing teams: brainstorm on Monday, run a few variants on Tuesday, finalize on Wednesday — the pace is fully under control and the cost is low. The caveat is that Remotion is free for individuals and small teams, while corporate commercial use requires attention to its license.

It Has Its Limits Too

Of course, it is not omnipotent. The whole process is code-driven, so if you want to deeply tweak the camera rhythm, you had better understand a bit of TypeScript / React. It is also better suited to promo videos, feature demos, and brand films for web and desktop products, rather than films with live-action footage or heavy narrative. In addition, most audio assets come from Mixkit’s free commercial library, so before an official release you should still verify licensing and replace the demo screenshots in the template.

How to Get Started

Installation is simple. In Claude Code / Codex, just say “Install this skill for me: https://github.com/Vincentwei1021/video-shotcraft”, or use npx skills add Vincentwei1021/video-shotcraft. You can also git clone it and symlink it into your agent’s skills directory.

Using it is even more conversational: tell the agent “Use video-shotcraft to make a promo video for my desktop product, with the deck-deal-flyin and row-embed cards,” and it gets to work. When you do not specify shot cards, it first recommends built-in templates for you to choose from.

Closing Thoughts

Ultimately, video-shotcraft compresses “cinematic product video” — a task that used to be piled up from people and experience — into a reusable, reproducible agent workflow. It may not replace a professional studio, but it does cut the barrier down to a height an independent developer can reach. If you happen to have a product to promote, it is worth handing it over for a try — the worst outcome is simply accumulating a few more sets of ready-to-use shot assets.

Repository (including the online gallery and showcase videos): https://github.com/Vincentwei1021/video-shotcraft

Online gallery: https://vincentwei1021.github.io/video-shotcraft/