This project provides an open-source, automated tool for creating short-form videos for platforms like TikTok and Instagram Reels, serving as a free alternative to expensive API services and GPU-intensive generation tools. It targets content creators and developers looking to automate video production from text inputs.

How It Works

The tool orchestrates several components to generate videos: Kokoro TTS converts input text to speech, Whisper.cpp generates accurate captions from the audio, Pexels API provides background video clips based on search terms, and Remotion composes these elements into a final video with styled captions. This approach allows for programmatic video creation without requiring users to manually source assets or perform complex editing.

Quick Start & Requirements

• Install/Run: Docker is recommended.
  • docker run -it --rm --name short-video-maker -p 3123:3123 -e LOG_LEVEL=debug -e PEXELS_API_KEY=<your_pexels_api_key> gyoridavid/short-video-maker:latest-tiny
• Prerequisites:
  • NPM (if not using Docker)
  • Docker (recommended)
  • Pexels API Key (free)
  • Linux (Ubuntu ≥ 22.04) or macOS. Windows is not supported.
  • Minimum 3GB RAM (4GB recommended), 2 vCPUs, 5GB disk space.
  • For GPU acceleration (Whisper.cpp), an NVIDIA GPU and CUDA are required (use latest-cuda Docker image).
• Resources: Official Docker images and documentation are available.

Highlighted Details

• Supports both REST API and Model Context Protocol (MCP) for integration with AI agents like n8n.
• Offers customization for music genre/mood, voice, caption styling, and video orientation.
• Includes a basic health check and status endpoints for video generation jobs.
• Provides endpoints to list available voices and music tags.

Maintenance & Community

The project was open-sourced by the AI Agents A-Z YouTube Channel. Contributions are welcome via Pull Requests.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Permissive MIT license allows for commercial use and integration into closed-source projects.

Limitations & Caveats

• Currently supports English voiceovers only due to Kokoro.js limitations.
• Background videos are sourced exclusively from Pexels.
• Does not support stitching user-provided images or videos.
• Remotion rendering is CPU-intensive; GPU acceleration is limited to Whisper.cpp.