media-mcp

Skill Pilot Media Server

An all-in-one Media Server for Skill Pilot AI Agent https://skill-pilot.ai. Includes ComfyUI for image/video generation, IndexTTS for text-to-speech, MuseTalk for talking video and live avatar, and SongBloom for music creation.

Included Models and Tools

How to Use

docker pull skillpilot/media-server:latest
docker run --gpus all \
  -p 18188:8188 -p 17860:7860 -p 18080:8080 \
  -p 13478:3478 -p 13478:3478/udp \
  -p 15349:5349 -p 15349:5349/udp \
  skillpilot/media-server:latest

When the container is running, it will automatically download the modeles by comand line below:

huggingface-cli download skill-pilot/media-mcp --local-dir /home/ubuntu/workspace/models

Use command below to check the downloading process in the container:

tmux attach -t download -r

You can use the services until the downloading process is completed. The models will be stored in the /home/ubuntu/workspace/models directory inside the container.

For how to use the media mcp server, please check at https://skill-pilot.ai, and find our Discord server invite link at https://skill-pilot.ai and join us for support.

The image is built on runpod/pytorch:2.4.0-py3.11-cuda12.4.1-devel-ubuntu22.04, ensuring compatibility with a wide range of NVIDIA GPUs and CUDA versions.

Tested on NVIDIA RTX 2060 GPU (12 GB VRAM) - MuseTalk live avatar can have real-time performance at 12 FPS.

Supports local GPU acceleration with NVIDIA drivers on the host machine or Runpod Cloud environments.

Free community support — no account required.

Join our Discord server for help, setup, tips, and updates. Find the invite link at https://skill-pilot.ai.

License

Free and open source https://skill-pilot.ai · MIT License

All models included in this repository are licensed under their respective licenses. Please refer to the individual model documentation for details.