sertdfyguhi / onGAU

onGAU: Oversimplified Neural Generative AI UI

onGAU: Oversimplified Neural Generative AI UI

A very simple AI image generator UI interface built with Dear PyGui and Diffusers.

Installation

1. Clone this repository.

git clone https://github.com/sertdfyguhi/onGAU.git

2. Run install.py with python to setup onGAU.

python3 install.py

3. Run run.sh for MacOS / Linux or run.bat for Windows to start onGAU.
4. Happy generating!

Scripts

pnginfo.py:
   Shows the png metadata (prompt, negative prompt...) of a png file.

esrgan.py:
   CLI to upscale an image using ESRGAN.

Todo

• Show total generation time
• Save and load prompts and config
• Add img2img pipeline
• Add LPW stable diffusion pipeline
• Add textual inversion model loading
• Add Clip Skip parameter
• Add model CPU offloading
• Average step time
• Generation Progress ETA
• Load .ckpt in app
• Load .safetensors lora in app
• Implement lora correctly and fully
• Add tooltips
• Denoising strength
• Change embedding models and loras in app
• Rework and organize UI
• Merging models
• Add controlnet support
• Add super resolution (ESRGAN/SwinIR) support
  • Add ESRGAN support
  • Add SwinIR support
• Create an install script to easily install UI
• Interrupt generation process
• Code Optimization
  • Fix memory leak when Compel prompt weighting is enabled
  • Use better code to get image size
  • Sometimes after changing models inference is extremely slow
  • When switching pipelines inference gets extremely slow
  • First step takes extra time after ESRGAN upscale
  • Merging models makes inference unusably slow