DiffuGen Image Generator

Example Use Cases

• 1

  Generate images locally from natural language prompts inside IDE chat interfaces (Cursor, Windsurf, Roo Code, Cline) via MCP.

• 2

  A programmatic HTTP/OpenAPI endpoint to produce and retrieve generated images for integration into custom applications or OpenWebUI.

• 3

  Reproducible, GPU‑accelerated image generation with fine-grained control over models, sampling, seeds, dimensions, and negative prompts.

Description

DiffuGen is an MCP server and toolchain that enables local text-to-image generation using optimized stable-diffusion.cpp and multiple model families (Flux and Stable Diffusion). It integrates with MCP-compatible IDEs (Cursor, Windsurf, Roo Code, Cline) to allow image generation from the editor, and also exposes an OpenAPI/REST interface and OpenWebUI tooling for broader programmatic access. The project supports multi-GPU/CUDA acceleration, fine-grained parameter control (steps, CFG, sampling method, negative prompts), and configurable model and output paths for reproducible local workflows. It is aimed at enabling fast drafts and production-quality images without relying on external cloud APIs.