AIUpscaleGUI

• Using upscaling models for upscaling images.

Table Of Contents

• AIUpscaleGUI
  • Table Of Contents
  • Changelogs
  • Requirements
    • Check Tkinter Module
  • Usage
    • Folder Selection Note
    • Model Names And Attributes
  • Examples

Changelogs

• See the changelogs.md file

Requirements

• Tested in Windows OS environment.
• Git has to be in PATH variable.
• Tested with Python 3.12 and used Tkinter for GUI. As a note for Python 3.10.6 version Tkinter comes as default package. You can easily test it with these commands:

Check Tkinter Module

• From command line enter python command and press Enter button.
• Write import tkinter command and press Enter button. If there are no errors. You are ready to go.
• Write exit() command and press Enter button to exit.

Usage

• Open terminal or cmd and clone the repository and get into folder git clone https://github.com/alpertunga-bile/AIUpscaleGUI.git & cd AIUpscaleGUI
• Start the application with python start.py command. This command is going to look for 'venv' file for virtual environment. It is going to setup the dependencies (customtkinter) and start the application. After the first setup, it is just going to start the GUI application.
• Choose your input and output directories.
• Then select which model and attributes you want to use.
• After selecting them, press Initialize button and wait for Done !!! text under the button.

Folder Selection Note

• Recursive image file location support is added.

Model Names And Attributes

• Supported models and their supported scale values are listed below. Select one model from the table:

Model Name	Supported Scale
RealESRGAN_x4plus	4x
RealESRNET_x4plus	4x
RealESRGAN_x4plus_anime_6B	4x
RealESRGAN_x2plus	2x
2x-AnimeSharpV4_RCAN	2x
2x-AnimeSharpV4_Fast_RCAN_PU	2x
4x-UltraSharpV2	4x
2x-ModernSpanimationV1	2x
2x_Ani4Kv2_G6i2_Compact_107500	2x
4xmssim_drct-l_pretrain	4x
4xmssim_drct_pretrain	4x
4xmssim_span_pretrain	4x

• You can select bfloat16 float16 and float32 PyTorch data types.

Examples

• Prompts are parsed with Prompt Markdown Parser project.
• RealESRGAN_x4plus model, x4 scale and face enhancement are used.
• 512x512 images are created with NeverEndingDreamBakedVAE model, model hash is 64b14b6ca5. CodeFormer is used for face restoration.

512x512	2048x2048

512x512	2048x2048