sd.cpp.gui.wx

Cross platform wxWidgets based Stable Diffusion C++ GuI

MIT License

Features

• detect environment to which HW available for diffusion
• text2img generation
• controlnet support
• image2image generation
• upscaler with some limitations
• minimalistic model management, model info from CivitAi.com
• vae support
• presets management (save / restore some config from UI)
• generation queue & queue list
• progress bar in the queuelist, speed (it/s or s/it)
• restore last item from queue list, auto start last pending job
• clone queue item, copy generation data from queue item
• save generation metadata into images
• load generation data from images
• image drag & drop

Compatibility

Please see github releases for installers. These win packages are tested on Windows 11, .deb packages tested on Ubuntu 24

Use only at your own risk! These releases just previews, so crashes may occur. For bugreport, please use github Issuses.

Starting with selected backend

To select a backend on startup, add one from the following parameters:

• -cuda
• -avx
• -avx2
• -avx512
• hipblas
• vulkan

To allow to run in multi - instance mode, start with -allow-multiple argument. Otherwise only one instance is allowed to run in

Building on Linux / Windows

Requirements - Windows

• latest conan for package dependencies (winget install -e --id JFrog.Conan)
• latest cmake (winget install -e --id Kitware.CMake)
• a compatible compiler (Visual Studio Community 17 2022) (winget install --id=Microsoft.VisualStudio.2022.Community -e)
• a git client (winget install -e --id Git.Git)
• Optional NSIS for creating windows installer (winget install -e --id NSIS.NSIS)

Compile on Windows - pre-requirements

Open a developer terminal, then run:

git clone https://github.com/fszontagh/sd.cpp.gui.wx.git
mkdir build
cd build

Run conan to gather dependencies:

conan install ..

Requirements - Linux

• build tools

  apt install build-essential
  

• exiv2

  apt install libexiv2-dev
  

• OpenSSL:

  apt install libssl-dev
  

• libjpeg

  apt install libjpeg-dev
  

• libpng

  apt install libpng-dev
  

• libtiff

  apt install libtiff-dev
  

• libgtk-3

  apt install libgtk-3-dev
  

• libcurl

  apt install libcurl4-openssl-dev
  

• libsecret-1-dev

  apt install libsecret-1-dev
  

• libnotify-dev

  apt install libnotify-dev
  

• libwebkit2gtk-4.1-dev

  apt install libwebkit2gtk-4.1-dev
  

• libsdl2-dev

  apt install libsdl2-dev
  

Configure the project:

Run in the build folder:

cmake .. 

If you configure without any parameter, the sd.cpp libs will not be compiled. You need to specify the correct backend to compile. For CPU only, please checkout whitch cpu features are available on your platform. For AVX only:

cmake .. -DSD_AVX=ON

For AVX2 only:

cmake .. -DSD_AVX2=ON

For AVX512

cmake .. -DSD_AVX512=ON

For CUDA only:

cmake .. -DSD_CUBLAS=ON

For VULKAN only:

cmake .. -DSD_VULKAN=ON

For HIPBLAS only:

cmake .. -DSD_HIPBLAS=ON

It is possible to combine more parameters

cmake .. -DSD_CUBLAS=ON -DSD_AVX=ON -DSD_AVX2=ON -DSD_AVX512=OFF

Of course, for VGA backends, need to be install the required tools.

To compile, run:

cmake --build . --config Release

To create package:

cmake --build . --config Release --target package