Heavily opinionated custom fork of https://github.com/AUTOMATIC1111/stable-diffusion-webui
Fork is as close as up-to-date with origin as time allows
All code changes are merged upstream whenever possible
The idea behind the fork is to enable latest technologies and advances in text-to-image generation
Sometimes this is not the same as "as simple as possible to use"
If you are looking an amazing simple-to-use Stable Diffusion tool, I'd suggest InvokeAI specifically due to its automated installer and ease of use
Follow Development updates for daily updates on new features/fixes
- New installer
- Advanced CUDA tuning
Available in UI Settings - Advanced environment tuning
- Optimized startup and models lazy-loading
- Built-in performance profiler
- Updated libraries to latest known compatible versions
- Includes opinionated System and Options configuration
- Does not rely on
Accelerate
as it only affects distributed systems
Gradio web server will be initialized much earlier which model load is done in the background
Faster model loading plus ability to fallback on corrupt models - Uses simplified folder structure
e.g./train
,/outputs/*
,/models/*
, etc. - Enhanced training templates
- Built-in
LoRA
,LyCORIS
,Custom Diffusion
,Dreambooth
training - Majority of settings configurable via UI without the need for command line flags
e.g, cross-optimization methods, system folders, etc. - New logger
- New error and exception handlers
- Optimized for
Torch
2.0 - Runs with
SDP
memory attention enabled by default if supported by system
Note:xFormers
and other cross-optimization methods are still available - Auto-adjust parameters when running on CPU or CUDA
Note: AMD and M1 platforms are supported, but without out-of-the-box optimizations
Hand-picked list of extensions that are deeply integrated into core workflows:
- System Info
- ControlNet
- Image Browser
- LORA (both training and inference)
- LyCORIS (both training and inference)
- Model Converter
- CLiP Interrogator
- Dynamic Thresholding
- Steps Animation
- Seed Travel
- Multi-Diffusion Upscaler
- Includes updated UI: reskinned and reorganized
Black and orange dark theme with fixed width options panels and larger previews - Includes support for Gradio themes
Settings -> User interface -> UI theme
Link to themes list & previews: https://huggingface.co/spaces/gradio/theme-gallery
- Drops compatibility with older versions of
python
and requires 3.9 or 3.10 - Drops localizations
Fork adds extra functionality:
- New skin and UI layout
- Ships with set of CLI tools that rely on SD API for execution:
e.g.generate
,train
,bench
, etc.
Full list
- Install first:
Python & Git - If you have nVidia GPU, install nVidia CUDA toolkit:
https://developer.nvidia.com/cuda-downloads - Clone repository
git clone https://github.com/vladmandic/automatic
Run desired startup script to install dependencies and extensions and start server:
webui.bat
andwebui.sh
:
Platform specific wrapper scripts For Windows, Linux and OSX
Startslaunch.py
in a Python virtual environment (venv)
Note: Server can run without virtual environment, but it is recommended to use it to avoid library version conflicts with other applications
If you're unsure which launcher to use, this is the one you wantlaunch.py
:
Main startup script
Can be used directly to start server in a manually activatedvenv
or to run server withoutvenv
setup.py
:
Main installer, used bylaunch.py
Can also be used directly to update repository or extensions
If running manually, make sure to activatevenv
first (if used)webui.py
:
Main server script
Any of the above scripts can be used with --help
to display detailed usage information and available parameters
For example:
webui.bat --help
Full startup sequence is logged in setup.log
, so if you encounter any issues, please check it first
The launcher can perform automatic update of main repository, requirements, extensions and submodules:
- Main repository:
Update is not performed by default, enable with--upgrade
flag - Requirements:
Check is performed on each startup and missing requirements are auto-installed
Can be skipped with--skip-requirements
flag - Extensions and submodules:
Update is performed on each startup and installer for each extension is started
Can be skipped with--skip-extensions
flag - Quick mode: Automatically enabled if timestamp of last sucessful setup is newer than actual repository version or version of newest extension
This repository comes with a large collection of scripts that can be used to process inputs, train, generate, and benchmark models
As well as number of auxiliary scripts that do not rely on WebUI, but can be used for end-to-end solutions such as extract frames from videos, etc.
For full details see Docs