🌌 MediaMCP

MediaMCP is a powerful Model Context Protocol (MCP) server designed for high-performance, local AI media generation. It provides a seamless interface for LLMs to generate and edit images, as well as compose and cover music using local backend services.

---

✨ Features

• 🖼️ Image Generation: Generate high-fidelity images using the Flux model via stable-diffusion.cpp.
• 🎨 Image Editing: Perform image-to-image transformations and edits with text guidance.
• 🎵 Music Composition: Create complete songs with vocals using ACE Step.
• 🎤 Cover Generation: Transform existing audio into new styles or voices (voice conversion).
• 🚀 FastMCP Powered: Built on the modern FastMCP framework for low-latency, scalable tool execution.
• 🔌 Standardized Interface: Exposes a clean API for any MCP-compliant client (like Claude Desktop or custom agents).

---

🛠️ Installation

Prerequisites

• Python 3.11 or higher
• Access to local media generation backends:
  • stable-diffusion.cpp (or compatible API) for images.
  • ACE Step CPP (or compatible API) for music.

Setup

1. Clone the repository:

   git clone https://github.com/haervwe/media-mcp.git
   cd media-mcp

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -e .

4. Configure environment variables: Copy the example environment file and fill in your API keys and endpoints:

   cp .env.example .env

   Edit .env to match your local setup.

---

🚀 Usage

Start the MediaMCP server:

python -m media_mcp.server

The server will start on the host and port specified in your .env (default: 0.0.0.0:8080) using the streamable-http transport.

---

🧰 Tools Provided

Tool	Description	Key Parameters
generate_image	Generate a new image from text.	prompt, format
edit_image	Edit an existing image.	image (path/b64), prompt, format
generate_song	Create a full song with vocals.	prompt, lyrics, language, key
generate_cover	Create a cover of an audio file.	audio (path/b64), style_prompt, strength

---

⚙️ Configuration

MediaMCP is highly configurable via environment variables in the .env file:

• IMAGE_API_BASE_URL: Endpoint for your image generation service.
• MUSIC_API_BASE_URL: Endpoint for your music generation service.
• ASSETS_DIR: Local directory where generated media files are stored.
• RESPONSE_FORMAT: Choose between path (file system path) or base64 (inline content).
• REQUEST_TIMEOUT: Timeout for long-running generation tasks (default: 300s).

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Developed with ❤️ for the AI community.