What it does
Stable Diffusion web UI is an open-source project that provides a web interface for Stable Diffusion, implemented with the Gradio library. It is published as a GitHub repository under the AGPL-3.0 license. Beyond the original txt2img and img2img modes, it offers inpainting, outpainting, color sketch, a prompt matrix, attention weighting in prompts, negative prompts, saved styles, variations, a highres fix and batch processing. An Extras tab collects face restoration and several neural-network upscalers. Generation parameters are saved in the image file and can be restored by dragging the image back into the interface. Training features include textual inversion embeddings, hypernetworks and LoRAs, with a training tab and image preprocessing. A checkpoint merger, checkpoint reloading on the fly, safetensors loading, and community extensions and custom scripts widen what it can do. The README mentions support for 4GB video cards, with reports of 2GB working. Installation is by release package or scripts on Windows, Linux and Apple Silicon. NVIDIA GPUs are the recommended path, with instructions linked for AMD, Intel and Ascend NPU hardware, and Python and git must be installed for the install script.
Capabilities
- txt2img and img2img with inpainting, outpainting and color sketch
- Prompt tools such as attention weighting, negative prompts, prompt matrix and saved styles
- Extras tab with face restoration and several upscalers
- Textual inversion, hypernetworks and LoRA support, with a training tab
- Checkpoint merger and checkpoint reloading on the fly
- Extensions and custom scripts from the community
When to use it
- Generating and editing images with Stable Diffusion models from a browser interface
- Fixing or extending part of an image with inpainting or outpainting
- Comparing prompt and parameter changes with X/Y/Z plots and prompt matrices
- Training textual inversion embeddings or merging checkpoints
