claude-code-video-toolkit: Claude Code from script to MP4

npx skills add digitalsamba/claude-code-video-toolkit --skill remotion

Ask an agent for an explainer video and it stalls in the middle. It can write the script, but voiceover, music, artwork, editing and rendering are spread across six or seven tools, and the progress is gone when the session ends. claude-code-video-toolkit, from Digital Samba (MIT, created December 9, 2025, about 2.2k stars as of October 8, 2026), puts Remotion rendering, AI voiceover, music, images, video clips and slash commands like /video into one repo, so Claude Code can take a project from concept to a finished MP4.

The order it works in

Open Claude Code in the repo and run /setup. By the README’s account it takes about five minutes to configure a cloud GPU, file storage and a voice. After that /video starts a project. It scans projects/*/project.json and lists each project’s phase and progress; with none, it asks you to pick a template (sprint-review, product-demo, or the vertical concept-explainer-short) and a brand, then gathers content, splits scenes and drafts the voiceover script.

A project moves through a fixed chain in project.json: planning, assets, review, audio, editing, rendering, complete. Each time you return, the command compares what the plan says with what is actually on disk, then regenerates a project-level CLAUDE.md that tells the agent what to do next. Once assets are in, it hands off to /scene-review, which walks every scene in Remotion Studio before any voiceover is generated.

What it bundles

• Skills: remotion, ffmpeg, elevenlabs, ltx2, ideogram4, acestep, qwen-edit, moviepy, playwright-recording and more, each covering one kind of job.
• Slash commands: /setup, /video, /scene-review, /design, /brand, /generate-voiceover, /record-demo, /publish and a handful more.
• Cloud GPU tools: Qwen3-TTS for voice, FLUX.2 for images, ACE-Step for music, LTX-2.3 for video clips, switched with --cloud modal or --cloud runpod.
• Transitions: besides the official ones, seven custom effects including glitch(), rgbSplit(), lightLeak() and clockWipe().

The README’s own reference prices (the author’s figures, not independent tests): about $0.01 for a voiceover and about $0.23 for an AI video clip. To see it work first, npm install && npm run render in examples/hello-world produces an MP4 with no API keys.

Who it’s for

Individuals and small teams who make product demos, sprint reviews or vertical explainers, and want the agent to remember project state, can clone the repo and go. Compared with OpenMontage, the difference is that you deploy the models yourself: the AI steps run on your Modal or RunPod account, pay as you go, with no closed video platform in the loop. The price of that is managing a cloud account. For a narrower single-purpose skill, see srt-whiteboard-animation.

Keep the install path straight. npx skills add https://github.com/digitalsamba/claude-code-video-toolkit --skill remotion installs only the remotion skill: the transitions library, shared components and project conventions such as 30 fps. It does not include /video or the templates, so for the full workflow you clone the whole repo. The repo also ships remotion-official, synced from Remotion’s own skills repo, for core usage.

Three honest limits. First, voice, images, music and video clips all depend on third-party cloud GPUs: Modal’s Starter plan gives $30 a month of free compute but needs a payment method, and RunPod bills per second. Second, Remotion has its own license, and companies may need to buy one for commercial use; the repo’s MIT license doesn’t cover it. Third, the author says in the README that the main use case is sprint review videos. Other genres rely on Claude Code adapting a template on the fly, and quality there isn’t guaranteed.