NMKD Stable Diffusion GUI
Local AI image generator for Windows 10/11 with Stable Diffusion, inpainting and LoRA training. Open source under GPL-3.0, with NVIDIA and AMD GPU support.
NMKD Stable Diffusion GUI is a local AI image generator for people who want to create and edit images on a Windows PC. It combines Stable Diffusion generation with inpainting, LoRA training and image post-processing in a desktop interface. It's open source under GPL-3.0.
You can generate images from text or use an existing image as a starting point. Negative prompts, weighted phrases and textual inversion embeddings give you control over the result, while LoRA support lets you use specialized models. Batch prompts and a queue help when you want to compare variations. Prompt history stores earlier prompts for reuse.
For image editing, InstructPix2Pix accepts written instructions. RealESRGAN handles upscaling, and GFPGAN or CodeFormer can restore faces. The generator also supports tileable textures, mirrored images and a high-resolution fix to reduce repeated patterns in larger outputs.
The app uses a customized fork of InvokeAI for its main Stable Diffusion backend. An ONNX backend supports AMD and other DirectML-capable GPUs, though it runs more slowly and has fewer features. InstructPix2Pix requires NVIDIA hardware.
It runs on 64-bit Windows 10 and 11. The stated minimum is an NVIDIA GPU with 4 GB VRAM and 8 GB system RAM; the DirectML route needs 8 GB VRAM. CPU generation is available but very slow. Built-in model tools can merge checkpoints, reduce their size and convert weights between PyTorch, Diffusers, ONNX and SafeTensors formats. Downloaded checkpoints, adapters and post-processing models have their own license terms; the GUI’s GPL-3.0 license covers its software.
