Stable Diffusion web UI

A browser-based interface for running Stable Diffusion image-generation models locally. The application is built around the Gradio framework, with a FastAPI HTTP API on the side and a large extension/script ecosystem layered on top.

What this project is

stable-diffusion-webui (often called "AUTOMATIC1111" or "A1111" after its original author) is the most widely deployed open-source frontend for Stable Diffusion. It bundles:

• A txt2img and img2img workflow with prompt parsing, attention syntax, X/Y/Z grids, and infinite-token prompts.
• An extras/post-processing pipeline (face restoration, upscaling, image-to-image alternatives).
• A training tab for textual inversion embeddings and hypernetworks.
• A built-in Lora extension and an extensions tab that downloads, installs and updates third-party plugins from a registry index.
• A REST API at /sdapi/v1/* that mirrors most UI features and is documented automatically via OpenAPI.

The code lives almost entirely in modules/ and extensions-builtin/; the rest of the repository is launch wrappers (launch.py, webui.sh, webui.bat), Gradio JavaScript glue (javascript/), CSS, and YAML configs for the underlying ldm/Stable Diffusion model.

What this wiki covers

This wiki is generated from the source tree at commit 82a973c0. It is a codebase wiki — it explains how the program is structured, where the main extension points are, and what each module does. End-user documentation (model recommendations, prompt tutorials, command-line tricks) lives on the GitHub wiki.

You want to…	Start here
Understand the overall architecture	overview/architecture.md
Run the project locally	overview/getting-started.md
Look up a project-specific term	overview/glossary.md
See how the codebase has evolved	../lore.md
Read the source-tree statistics	../by-the-numbers.md
Modify a subsystem	the systems/ section
Implement a new feature for users	the features/ section
Write an extension or script	systems/scripts-and-extensions.md and systems/script-callbacks.md
Drive the API	api/index.md

Quick orientation

launch.py                # bootstraps the venv, installs deps, calls webui.py
webui.py                 # api_only() and webui() entry points
modules/                 # ~120 Python modules — the bulk of the application
modules/api/             # FastAPI routes and Pydantic models
modules/processing.py    # the txt2img / img2img generation pipeline
modules/sd_models.py     # checkpoint loading and switching
modules/scripts.py       # the script and extension registry
modules/ui.py            # top-level Gradio UI assembly
extensions-builtin/      # Lora, LDSR, SwinIR, ScuNET, hypertile, etc.
javascript/              # browser-side helpers loaded into the Gradio page

The application is single-process. Generation requests are serialised by a global queue_lock defined in modules/call_queue.py, so only one image is being produced on the GPU at a time even though the API and the UI both feed into the same process.