Part of the StudioMeyer MCP Stack — Built in Mallorca 🌴 · ⭐ if you use it

mcp-video

License Last commit

**Cinema-grade video production for AI agents.**

8 MCP tools for recording, editing, effects, captions, TTS, and smart screenshots.
Built on ffmpeg and Playwright. Works with any MCP client.

Features · Quick Start · Examples · Architecture

A note from us

We have been building tools and systems for ourselves for the past two years. The fact that this repo is small and has few stars is not because it is new. It is because we only just decided to share what we have built. It is not a fresh experiment, it is a long story with a recent commit.

We love building things and sharing them. We do not love social media tactics, growth hacks, or chasing stars and followers. So this repo is small. The code is real, it gets used, issues get answered. Judge for yourself.

If it helps you, sharing, testing, and feedback help us. If it could be better, an issue is more useful. If you build something with it, tell us at hello@studiomeyer.io. That genuinely makes our day.

From a small studio in Palma de Mallorca.

Features

Tool	Operations	Description
`video_record`	cinema, scroll, multi-device	Record websites at 60fps with frame-by-frame capture
`video_edit`	speed, crop, reverse, keyframe, pip	Edit clips with zoom/pan, PiP, slow-mo
`video_color`	grade, effect, lut, chroma	Color grading, 22 LUT presets, green screen
`video_audio`	extract, music, ducking, mix, voice	Audio extraction, mixing, 9 voice effects
`video_text`	subtitles, caption, overlay, animate	Burn SRT, Whisper auto-caption, 15 text animations
`video_compose`	concat, intro, social, beat-sync, templates	Join clips, social format conversion, beat sync
`video_speech`	generate, voices, narrated	ElevenLabs/OpenAI TTS, full narrated videos
`video_screenshot`	capture, detect	Element-aware screenshots, page feature detection

Highlights

60fps frame-by-frame capture — Playwright screenshots every frame, ffmpeg encodes. Zero frame drops.
Cinema easing curves — 16 easing options including cinematic and showcase for buttery smooth scrolling.
Smart screenshots — Auto-detects 15+ UI elements (chat widgets, pricing sections, booking forms, etc.).
Narrated videos — Provide a URL + script, get a professional video with synchronized AI voiceover.
22 LUT presets — Film-grade color grading (teal-orange, noir, vintage, cyberpunk, etc.).
Social format export — One-click conversion to Instagram Reel, TikTok, YouTube Short, LinkedIn.
Dual transport — Stdio (default) or HTTP mode for persistent microservice deployment.

Prerequisites

Node.js >= 18
ffmpeg and ffprobe (validated on startup, cross-platform)
Playwright browsers (npx playwright install chromium)
Optional: ELEVENLABS_API_KEY for ElevenLabs TTS
Optional: OPENAI_API_KEY for Whisper captions and OpenAI TTS

If ffmpeg lives outside PATH, set FFMPEG_PATH and FFPROBE_PATH to the absolute binary paths. Both env vars are honoured at startup AND at every runtime spawn site.

Quick Start

With Claude Code (stdio)

config.json

{
  "mcpServers": {
    "video": {
      "command": "npx",
      "args": ["-y", "mcp-video"]
    }
  }
}

With npx

Terminal

npx mcp-video

From source

bash

git clone https://github.com/studiomeyer-io/mcp-video.git
cd mcp-video
npm install
npx playwright install chromium
npm run build
npm start

HTTP mode

bash

# Start as HTTP microservice
npx mcp-video --http --port=9847

# Or via environment variables
MCP_HTTP=1 MCP_PORT=9847 npx mcp-video

Configuration

Environment Variable	Default	Description
`VIDEO_OUTPUT_DIR`	`./output`	Directory for generated files
`FFMPEG_PATH`	—	Absolute path to `ffmpeg` binary if not on `PATH`
`FFPROBE_PATH`	—	Absolute path to `ffprobe` binary if not on `PATH`
`ELEVENLABS_API_KEY`	—	ElevenLabs TTS API key
`OPENAI_API_KEY`	—	OpenAI API key (Whisper + TTS)
`MCP_HTTP`	`false`	Enable HTTP transport
`MCP_PORT`	`9847`	HTTP port
`MCP_HOST`	`127.0.0.1`	HTTP bind address
`MCP_VIDEO_DEBUG`	`false`	Enable debug logging
`MCP_VIDEO_ALLOW_INTERNAL`	`false`	Set to `1` to allow URLs that resolve to localhost / private / metadata IPs. Local dev only — leave unset in production (SSRF guard).

What You Can Build

Use Case	Tools Used	Output
Product demo video	`video_record` → `video_text` → `video_audio`	60fps website recording + auto-captions + background music
Social media clips	`video_record` → `video_compose`	Record once → export to Instagram Reel, TikTok, YouTube Short
Narrated explainer	`video_speech` → `video_color`	AI voiceover + cinematic color grade
Before/after comparison	`video_screenshot` → `video_edit`	Smart element screenshots + PiP composition
Automated QA	`video_record` + `video_screenshot`	Record user flows + screenshot specific elements

Usage Examples

Record a website

Code

Use video_record with type "cinema" to record https://example.com
with a smooth scroll and hover over the navbar.

Create a narrated explainer video

Code

Use video_speech with type "narrated" to create a narrated video of
https://example.com with these segments:
1. "Welcome to our homepage" — pause on hero section
2. "Check out our features" — scroll to features
3. "Get started today" — hover over CTA button

Auto-caption a video

Code

Use video_text with type "caption" to add auto-generated captions
to /path/to/video.mp4

Export for social media

Code

Use video_compose with type "social-all" to convert
/path/to/video.mp4 to all social media formats.

Smart screenshot

Code

Use video_screenshot with type "capture" to screenshot the chat widget
and pricing section on https://example.com

Architecture

Code

src/
  server.ts            Entry point, 8 consolidated MCP tools
  lib/                 Logger, types, dual transport
  handlers/            Tool handlers (video, editing, post-production, tts, screenshots)
  schemas/             JSON Schema definitions for legacy tool format
  tools/
    engine/            Core engines
      capture.ts       Frame-by-frame recording (Playwright → PNG → ffmpeg)
      encoder.ts       ffmpeg encoding pipeline
      scenes.ts        Scene execution (scroll, hover, click, type, wait)
      cursor.ts        Visible cursor simulation
      smart-screenshot.ts  Element-aware screenshot engine
      tts.ts           ElevenLabs + OpenAI TTS with fallback
      narrated-video.ts    Full narration pipeline
      social-format.ts     Social media format conversion
      concat.ts        Video concatenation with transitions
      lut-presets.ts   22 cinema LUT presets
      ...and more

Development

Terminal

npm run dev          # Start with tsx (hot reload)
npm run typecheck    # Type check
npm test             # Run tests
npm run check        # Verify ffmpeg/ffprobe installed

About StudioMeyer

StudioMeyer is an AI and design studio based in Palma de Mallorca, working with clients worldwide. We build custom websites and AI infrastructure for small and medium businesses. Production stack on Claude Agent SDK, MCP and n8n, with Sentry, Langfuse and LangGraph for observability and an in-house guard layer.

License

MIT

Credits

Built by StudioMeyer. Part of our open-source toolkit for AI-powered content creation.

ai-shield — LLM security middleware for TypeScript
agent-fleet — Multi-agent orchestration for Claude Code
darwin-agents — Self-evolving agent framework