MCP_server_local_image_and_music_generation

This repository provides a local MCP (Model Context Protocol) server exposing open source image and audio generation tools. Inference runs on your machine using fast and lightweight modern models.

• Image Generation: Driven by Segmind SSD-1B (A distilled SDXL model providing 1024x1024 high quality with half the size of SDXL). Batched prompts are supported.
• Audio Generation: Driven by AudioLDM 2 for generating music, sound effects, and speech-style prompts.
• StdIO MCP server for integration with MCP-compatible clients.
• Local-first workflow (models downloaded on first run, outputs saved locally).
• Unified, clean dependencies using Hugging Face diffusers.

• Model: segmind/SSD-1B
• Why: Distilled version of Stable Diffusion XL (SDXL). It achieves massive quality leaps in prompt adherence, lighting, and anatomy, and it can accurately generate text—all while being 50% smaller and 60% faster than SDXL.
• Download Size: ~2.5GB
• VRAM Requirement: ~4GB+ for optimal hardware generation.

• Model: cvssp/audioldm2
• Why: Replaces earlier sequential generators like MusicGen. AudioLDM 2 gracefully handles a wide range of tasks—including complex music, environmental sound effects, and speech synthesis—in a single streamlined model.
• Download Size: ~1.1GB
• VRAM Requirement: ~2GB+

• mcp_server/main.py: MCP tool definitions and input validation
• image_gen/image_generator.py: SSD-1B pipeline and batched image generation
• audio_gen/audio_generator.py: AudioLDM2 loading and audio generation
• config.py: defaults and limits (including image batch limit)

• Python 3.8+
• Recommended for image generation: CUDA GPU with 6GB+ VRAM
• Recommended RAM: 16GB+
• For audio proces