Stable Diffusion Image Generation

This wraps HuggingFace's Diffusers library for running Stable Diffusion locally, which means you need GPU access but get full control over the generation pipeline. It covers text-to-image, img2img, inpainting, and ControlNet for spatial conditioning like edge maps or pose skeletons. The documentation includes scheduler comparisons (DPMSolver gets you quality in 15-25 steps instead of 50) and walks through SDXL and LoRA adapters. Use this when you want to generate images programmatically without API costs or need fine-grained control over the diffusion process. If you just need images and don't want to manage GPU inference, stick with DALL-E 3 or Midjourney instead.

Install to Claude Code

npx -y skills add orchestra-research/ai-research-skills --skill stable-diffusion-image-generation --agent claude-code