MangaJaNaiConverterGui

• Fix error upscaling when the username contains a space

• Fix additional errors upscaling grayscale images, regression from v1.1.5 (https://github.com/the-database/MangaJaNaiConverterGui/issues/33)

• Fix error upscaling grayscale images, regression from v1.1.5 (#33)

• Allow selection of No Model to skip running a model (#19)
• Fix console button cut off (#20)
• Fix handling of filenames containing multiple dots (#11)
• Save grayscale images in grayscale colorspace for applicable formats
• Optimize performance for upscaling grayscale images

• Fix NoneType bug when upscaling images which don't match any chains
• Add grayscale detection threshold option

• Test fix for batch upscale
• Show Python environment details in app settings

• Temporarily force torch 2.3.1 due to fbgemm.dll error on windows

• Fix upscaling errors when installing to a path containing spaces

• Redesigned app structure to separate Python environment and MangaJaNai models from the GUI, which are now downloaded and set up when launching the GUI for the first time. This design is inspired by chaiNNer.
• New design makes installing updates to the GUI much faster and more convenient - the Python environment and models don't need to be redownloaded for each update:
  • New Version (1.1.0): 17.4 MB filesize for setup.exe
  • Old Version (1.0.0): 1.8 GB filesize for setup.exe
• New design improves app startup time significantly:
  • New Version (1.1.0): 1.0 seconds to launch
  • Old Version (1.0.0): 5.5 seconds to launch

• Add all 14 MangaJaNai V1 models
• Simplify main upscale settings
• Add advanced options to be able to select model depending on image resolution and other conditions
• Add custom workflows
• Add support for loading and saving AVIF format images

NOTE: The setup exe may by falsely flagged by anti virus software due to not being code signed (https://github.com/the-database/MangaJaNaiConverterGui/issues/9), this is being investigated. In the meantime, if you run into this issue, try using the portable version instead, and run the MangaJaNaiConverterGui.exe inside the current directory instead of the outer one.

• Fix failure to open some images (#12)
• Fix color images being detected as grayscale (#7)
• Fix auto update (#10). Auto update remains broken in all releases prior to 0.0.13, please update to 0.0.13 manually if you are on an older version.

NOTE: The setup exe may by falsely flagged by anti virus software due to not being code signed (#9), this is being investigated. In the meantime, if you run into this issue, try using the portable version instead, and run the MangaJaNaiConverterGui.exe inside the current directory instead of the outer one.

• Add MangaJaNai Grayscale 1200 V1 models for grayscale manga pages:
  • 4x_MangaJaNai_Grayscale_1200_V1_ESRGAN_70k - Best quality model with balanced performance, recommended for most users. The 1200 model gives best results when the manga page is around 1200px tall. The Resize Height Before Upscale setting can be used to resize the image to 1200px before upscaling to improve results.
  • 4x_MangaJaNai_Graycsale_1200_V1_OmniSR_100k - Lower quality model, recommended for low end GPU or CPU users. The 1200 model gives best results when the manga page is around 1200px tall. The Resize Height Before Upscale setting can be used to resize the image to 1200px before upscaling to improve results.
• Fix grayscale image detection incorrectly detecting color images as grayscale (#7)
• Fix default app window resolution being larger than screen resolution (#6)

NOTE: The setup exe may by falsely flagged by anti virus software due to not being code signed (#9), this is being investigated. In the meantime, if you run into this issue, try using the portable version instead.

• Add IllustrationJaNai models for color images including manga covers and color illustrations, digital art, visual novel art, artbooks, and more. Add ESRGAN version of 4x_MangaJaNai_V1RC34 for grayscale manga images. The following models were added:
  • 4x_IllustrationJaNai_V1_DAT2_190k - Maximum quality but very slow, only recommended for use with high end GPUs.
  • 4x_IllustrationJaNai_V1_ESRGAN_135k - High quality with balanced performance, recommended for all users without a high end GPU
  • 4x_MangaJaNai_V1RC34_ESRGAN_760k - ESRGAN version of the RC34 OmniSR model released above. Halftone handling, text, lines, and artifact handling are significantly improved, but the model runs slower than the OmniSR model
• Set CPU mode by default for users with no GPU
• Fix bug where upscaling a folder of archives failed

• update chaiNNer backend to 20240215
• add chaiNNer backend options, including GPU selection, CPU mode (#1 ), and fp16 option (#3)
• add resize to width and upscale tile size settings (#4)
• fix app icon

• include imagemagick lib to fix error during upscale

• add installer with auto update

• add estimated time remaining and estimated finish time
• make folder dialogs default to currently selected folder

• reorganize input and output fields
• hide additional settings by default
• fix upscaling error on 1 channel images
• add support for rar/cbr

• add status bar and progress bars
• make console view togglable and hide by default
• prepopulate model fields