The GTK Stable Diffusion is a very simple text-to-image tool. The tool is based on GTK UI framework and Diffusers' Stable Diffusion text-to-image pipeline.

GTK Stable Diffusion aims just for fun. This means the tool is more focusing on spells prompts than many small adjustments and random generations.

Installation

GTK Stable Diffusion is easily installable via pip command:

pip install gtk_stable_diffusion

or

pip install git+https://github.com/nazodane/gtk_stable_diffusion.git

Note that the model weights are automatically downloaded via BitTorrent magnet link using libtorrent and the model weights are automatically converted from original Stable Diffusion format to Diffusers format in the first launch of the tool.

Usage

~/.local/bin/gtk-stable-diffusion

Note: ctrl+space will show the candidates of synonyms for prompt and antonyms for negative prompt

Requirements

• Linux
• Python 3.10 or later
• CUDA 11.7 or later
• DRAM 16GB or higher
• RTX 3060 12GB or higher (the VRAM usage is currently over 8GB!)

Recommendations

• Ubuntu 22.04 or later
• DRAM 32GB or higher
• NVMe SSD
• Faster non-restricted internet connections

License

GTK Stable Diffusion codes are under Apache License 2.0. This is because we almost depend on Diffusers.

GTK and its Python bindings are LGPL so we should carefully treat GTK-related codes.

Screenshot