> /create-video "30-second promo for a coffee shop: drone shot of the city,
warm voiceover, background jazz, animated shop name and address"
✓ script & scene plan ✓ voiceover (ElevenLabs) ✓ jazz (Eleven Music)
✓ city footage (Pexels) ✓ captions (Whisper) ✓ transitions
✓ live preview in Studio ✓ render → out/video.mp4 ✓ AI review → fixes
Remotion turns React into video, and Remotion's official Agent Skills teach your agent how to write it. Remotion Superpowers layers on top and gives your agent everything else a video needs. It can see (video analysis), speak (voiceovers), hear (music and SFX), source (stock footage and AI generation), caption, transition, and review its own render, then fix it.
Install
| Agent | Start with (full steps below) |
|---|
| Claude Code | /plugin marketplace add DojoCodingLabs/remotion-superpowers |
| Codex | codex plugin marketplace add DojoCodingLabs/remotion-superpowers |
| Gemini CLI | gemini extensions install https://github.com/DojoCodingLabs/remotion-superpowers |
| Cursor, Copilot, others | npx skills add DojoCodingLabs/remotion-superpowers |
Requires Node.js 22.18+ (it runs the TypeScript audio script directly), plus the official skills this plugin builds on, which /setup installs for you: npx skills add remotion-dev/skills and npx skills add elevenlabs/skills (and replicate/skills if you use Replicate). uv is only needed for the optional ElevenLabs MCP server.
Claude Code
/plugin marketplace add DojoCodingLabs/remotion-superpowers
/plugin install remotion-superpowers@remotion-superpowers
Then run /setup.
Codex
codex plugin marketplace add DojoCodingLabs/remotion-superpowers
codex plugin add remotion-superpowers@remotion-superpowers
Export your API keys in the shell you start Codex from; the plugin forwards them to the MCP servers. Then ask Codex to run the setup skill.
Gemini CLI
gemini extensions install https://github.com/DojoCodingLabs/remotion-superpowers
Gemini asks for each API key during install (leave the optional ones blank). Change one later with gemini extensions config remotion-superpowers "ElevenLabs API key".
Cursor, GitHub Copilot, and other agents
npx skills add DojoCodingLabs/remotion-superpowers
This installs the 15 skills into every agent it detects. Then add the servers from .mcp.json to your agent's MCP config and set the keys listed under each server's env_vars.
[!NOTE]
Built on the official skills. Remotion's own skills handle the markup, preview and rendering; ElevenLabs' handle voice, sound and music; Replicate's handle model choice. This plugin puts it all together for video: what to generate, where it goes, how it's timed, and the AI review loop.
Skills and MCP servers work in every agent above. The 3 subagents in agents/ and the hooks (API-key guard, captions tip) are Claude Code extras.
| Agent | Update |
|---|
| Claude Code | claude plugin marketplace update DojoCodingLabs/remotion-superpowers, then claude plugin update remotion-superpowers@remotion-superpowers, and restart the session |
| Codex | codex plugin marketplace upgrade |
| Gemini CLI | gemini extensions update remotion-superpowers |
| Agent Skills | npx skills update |
git clone https://github.com/DojoCodingLabs/remotion-superpowers.git
cd remotion-superpowers
# In Claude Code:
/plugin marketplace add .
/plugin install remotion-superpowers
Run bash scripts/ci/validate.sh before opening a PR. CI runs the same checks.
Quick start
1. Set up. Run /setup to check dependencies, create or detect a Remotion project, install Remotion's official skills, configure API keys, and verify the MCP servers.
2. Make a video. Run /create-video, or just describe it. Remotion Studio opens right away, so you can watch the video come together and steer:
"Create a 30-second promo video for a coffee shop. Include drone footage of a city, a warm voiceover, background jazz music, and animated text overlays with the shop's name and address."
3. Render and make it better. Ask for the MP4 when you're happy, or run /review-video: the agent renders, watches the result, lists what to fix, fixes it, and re-renders.
[!TIP]
Start your agent from the Remotion project root. Generated audio, images and clips land in that directory's public/.
What you get
| Capability | Powered by |
|---|
| 🗣️ | Voiceovers from a script | ElevenLabs (eleven_v4) |
| 🎵 | Background music from a mood | Eleven Music, cleared for commercial use |
| 🔊 | Sound effects from a description | ElevenLabs SFX + Remotion's hosted SFX |
| 📸 | Free HD/4K stock video and photos | Pexels |
| 🖼️ | AI images | KIE (Nano Banana Pro) + Replicate (FLUX.2, GPT Image and more) |
| 🎬 | AI video clips, text-to-video and image-to-video | KIE (Veo 3.1) + Replicate (Kling, Seedance, Wan) |
| 👁️ | Video understanding: describe footage, find the best moments | TwelveLabs (Pegasus) |
| 🔄 | AI review of your render, with timestamped fixes | TwelveLabs (Pegasus) |
| 📝 | TikTok-style word-by-word captions | Remotion's official captions (local Whisper, or ElevenLabs Scribe) |
| 🎯 | Scene transitions: fade, slide, wipe, flip, clock-wipe | @remotion/transitions (official skills) |
| 📱 | 9:16 short-form pipeline for TikTok, Reels, Shorts | Built-in presets |
| 🎲 | 3D scenes and product showcases | @remotion/three (official skills) |
| 📊 | Animated charts, dashboards, counters | SVG + D3 |
| ✨ | Light leaks, film grain, vignettes, color grading, Lottie | @remotion/effects (official skills) |
| 🚀 | Render on PR or on demand | GitHub Actions |
Commands
In Claude Code these are slash commands. In Codex, Gemini CLI and other agents they're skills with the same names: ask for one ("use the create-video skill") or just describe what you want.
| Command | What it does |
|---|
| Produce | /setup | First-run wizard: dependencies, project, keys, MCP check |
| /create-video | Full pipeline from a prompt to a rendered MP4 |
| /create-short | 9:16 vertical video with a hook-first structure and captions |
| Assets | /find-footage | Search and download Pexels stock footage |
| /generate-image | AI backgrounds, thumbnails, characters, art |
| /generate-clip | AI clips from text or from a still image |
| Audio | /add-voiceover | Generate narration and wire it into the composition |
| /add-music | Generate a track with fades and ducking |
| /transcribe | Audio or video to text with word-level timestamps |
| Polish | /add-captions | Animated word-by-word captions |
| /add-transitions | Professional scene transitions |
| Analyze | /analyze-footage | Understand existing footage with AI vision |
| /review-video | Render, review, fix, repeat |
How it works
Your prompt → /create-video (or /create-short)
│
├── 1. Studio → Remotion Studio opens as soon as the project runs; watch it build
├── 2. Concept → scenes, script, audio needs → you approve the plan
├── 3. Voiceover → per-scene ElevenLabs TTS, measured for timing
├── 4. Music → background track (Eleven Music)
├── 5. SFX → hosted or generated sound effects
├── 6. Footage → Pexels / AI images and clips (KIE, Replicate) / TwelveLabs analysis
├── 7. Visuals → Remotion markup via the official skills: scenes, transitions, effects
├── 8. Captions → word-level, TikTok-style
├── 9. Iterate → tweak live in Studio
├── 10. Render → when you ask: npx remotion render → MP4
└── 11. Review → /review-video → AI watches, suggests, iterates
API keys
| Service | Variable | Needed? | Cost | Unlocks |
|---|
| ElevenLabs | ELEVENLABS_API_KEY | ✅ Yes | Free tier; paid for commercial music | Voiceover, music, SFX, Scribe captions |
| KIE | KIE_API_KEY | ✅ Yes | Paid | AI images and clips (optional audio drafts) |
| TwelveLabs | TWELVELABS_API_KEY | ✅ Yes | Free tier | Footage analysis and AI review |
| Pexels | PEXELS_API_KEY | Recommended | Free | Stock photos and videos |
| Replicate | REPLICATE_API_TOKEN | Optional | Pay per use | Specific image and video models |
[!IMPORTANT]
KIE can also generate audio (Suno music, TTS, SFX) cheaply, but it resells those models without their owners' licence. Use KIE audio for drafts only; final audio comes from ElevenLabs.
Captions run locally with Whisper on a GPU (@remotion/whisper-webgpu), or through ElevenLabs Scribe when there isn't one.
MCP servers
The plugin configures 5 MCP servers for you, each pinned to a tested version:
- remotion-media (KIE): AI images and video clips.
- TwelveLabs (official): footage analysis and AI review.
- Pexels: free stock footage and photos.
- Replicate (official, optional): hundreds of image and video models.
- ElevenLabs (deprecated upstream, optional): quick interactive voice work. Voice, music and SFX normally go through the official ElevenLabs skills.
Works well with
Official tools from other providers, not bundled. Add them if you want more models:
- fal MCP: hundreds of image and video models
- Runway MCP: Runway's video models
- Gemini skills: Veo and Imagen through the Gemini API
- Codex: OpenAI's official Remotion plugin covers Remotion markup; this plugin adds the media on top.
Troubleshooting
You're on 2.1.0. Update to the latest release (see Updating under Install); 2.1.1 fixed the hooks schema.
- Check its key is set:
echo ${KIE_API_KEY:+set}. In Gemini CLI, use gemini extensions config instead, because extensions don't see your shell environment.
npx servers and the audio script need Node.js 22.18+. The optional ElevenLabs server (uvx) needs uv.- Restart your agent after changing keys. In Claude Code,
/mcp shows each server's status.
Generated files land in public/ of the directory your agent was started from. Start it from the Remotion project root.
Still stuck? Open an issue. Search the existing ones first; there's a good chance someone has hit the same thing.
remotion-superpowers/
├── .claude-plugin/ # Claude Code plugin + marketplace manifests
├── .codex-plugin/plugin.json # Codex plugin manifest
├── gemini-extension.json # Gemini CLI extension manifest
├── .mcp.json # 5 MCP server configurations (shared)
├── skills/ # 15 Agent Skills (SKILL.md, portable)
│ ├── setup/ # setup wizard
│ ├── create-video/ # full video production from a prompt
│ ├── create-short/ # TikTok/Reels/Shorts
│ ├── find-footage/ generate-image/ generate-clip/
│ ├── add-voiceover/ add-music/ transcribe/
│ ├── add-captions/ add-transitions/
│ ├── analyze-footage/ review-video/
│ ├── remotion-production/ # media orchestration (17 rules), on top of Remotion's official skills
│ └── setup-guide/ # setup & troubleshooting knowledge
├── agents/ # 3 Claude Code subagents
├── hooks/hooks.json # Claude Code hooks (API key checks, tips)
├── scripts/ # setup check, hook scripts, ci/validate.sh
├── CHANGELOG.md
└── README.md
Star history
If this saves you an afternoon of video editing, a ⭐ helps other builders find it.
Contributors
Thanks to everyone who has sent a fix, filed a sharp bug report, or forked this to build something of their own. New here? Read CONTRIBUTING.md. Better prompt flows, new production rules, and support for more agents are all welcome.
See CHANGELOG.md for what changed in each release.
Built by Dojo Coding
Dojo Coding is a LATAM-first tech education ecosystem. We build tools for developers and teach people to code.
More from Dojo Coding: CodeSensei (learn to code while you vibecode) · juan-workflow (dev lifecycle guardrails) · VibeCoding Bootcamp (structured curriculum with live mentors)
Free forever. Open source. MIT licensed.
