gyoridavid/short-video-maker

Creates short videos for TikTok, Instagram Reels, and YouTube Shorts using the Model Context Protocol (MCP) and a REST API.

1,268 ★407 forksTypeScriptUpdated 1y ago

What you need to know

Open-source automated short-form video creation tool: converts text input into narrated, captioned short videos by combining Kokoro text-to-speech, Whisper captions, Pexels background videos, music, and Remotion rendering. Exposes both an MCP (SSE) server for agents like n8n and a REST API for programmatic use.

Install

Docker (recommended): docker run -it --rm --name short-video-maker -p 3123:3123 -e LOG_LEVEL=debug -e PEXELS_API_KEY=<key> gyoridavid/short-video-maker:latest-tiny (or :latest, :latest-cuda for GPU)
npm/npx: requires Ubuntu >= 22.04 or macOS with ffmpeg and listed packages; node.js 22+

Usage

  • MCP server endpoints: /mcp/sse and /mcp/messages; tools: create-short-video and get-video-status
  • Use with n8n (local or Docker networking matrix documented) or any MCP client
  • REST API: POST /api/short-video with {"scenes": [{"text": "...", "searchTerms": ["river"]}], "config": {...}}, then GET /api/short-video/{id}/status and GET /api/short-video/{id} for the binary video
  • Configuration options: music mood (GET /api/music-tags), captionPosition, captionBackgroundColor, voice (GET /api/voices), orientation (portrait/landscape), musicVolume, paddingBack

Key features

  • Text-to-speech via Kokoro TTS, accurate captions via Whisper (whisper.cpp), background videos from Pexels, background music with genre/mood selection
  • Composes and renders with Remotion into professional-looking short videos with perfectly timed captions
  • Serves as both an MCP server and a REST server; example n8n workflows provided
  • Three Docker images: tiny (tiny.en whisper + q4 kokoro), normal (base.en + fp32), cuda (medium.en with GPU acceleration)
  • Web UI available at http://localhost:3123 for browser-based generation

Best for

Free, local, automated faceless short-video generation (n8n workflows or REST) without GPU-heavy or expensive third-party video APIs.

Caveats

  • English voiceover only (kokoro-js doesn't support other languages)
  • Background videos are sourced from Pexels only; cannot supply own videos/images or use video generation from images
  • Windows is NOT supported (whisper.cpp installation fails occasionally); Ubuntu >= 22.04 or macOS
  • Requires internet, a free Pexels API key, >= 3-4 GB RAM, >= 2 vCPU, >= 5 GB disk space
  • get-video-status tool is described as "somewhat useless" - use the REST API for status instead
Platforms: Linux · macOS · DockerClients: n8n · any MCP client
Documentation ↗

Reviewed 2026-08-07

Topics

faceless-videoinstagraminstagram-reelsinstagram-reels-automationmcpmodel-context-protocolreelsremotionshortsshorts-automateshorts-makertiktoktiktok-creatorvideo-creation-toolvideo-creatorvideo-generationvideo-generatorvideo-makeryoutubeyoutube-shorts
View on GitHub ↗
Stars
1,268★
Forks
407
Language
TypeScript
License
MIT
Created
2025-04-14
Last push
2025-06-21