Skip to content

Releases

2.6.0

Choose a tag to compare

@github-actions github-actions released this 12 Nov 06:17
7746e7f
  • Fix Fit to Display: the Tablet Device or Display list works again, now built into the app with over 800 tablets, e-readers, phones and monitors
  • Find devices by any part of their name, in any order, even with a typo
  • Type a resolution such as 1404x1872 for a device that isn't listed
  • List the connected monitors at their exact resolution
  • Keep a saved device's resolution even after it's renamed or removed from the list

2.4.2

Choose a tag to compare

@github-actions github-actions released this 08 Nov 03:22
93e1b1e
  • Pause now frees all of the GPU memory while paused

2.4.1

Choose a tag to compare

@github-actions github-actions released this 07 Nov 23:08
0ca47ef
  • Pause now frees the GPU memory while paused

2.4.0

Choose a tag to compare

@github-actions github-actions released this 07 Nov 21:39
e27614b
  • Pause and resume upscales, and continue one that didn't finish
  • Option to move the shared runtime and models to another drive (App Settings > Storage)
  • Removes the files MangaJaNai GUI 1.x left behind (several GB)
  • Scale mode now applies after Resize Before Upscale, instead of resizing the result back up
  • Files that would be saved under the same name get unique names instead of overwriting each other
  • Very wide or tall images such as banners are no longer stretched to giant sizes, and images too large for WebP or JPEG are shrunk to fit

2.3.0

Choose a tag to compare

@github-actions github-actions released this 07 Nov 08:37
37e7748
  • Support saving as JPEG XL, lossy or lossless
  • JPEG XL, TIFF and 7z/CB7 files can be picked as input and count toward a folder's progress

2.2.2

Choose a tag to compare

@github-actions github-actions released this 06 Nov 23:12
b77fb29
  • An update that brings a newer TensorRT now installs it

2.2.1

Choose a tag to compare

@github-actions github-actions released this 06 Nov 18:47
cfff920
  • The Linux AppImage is about 55 MB instead of 1.3 GB: on first launch it installs TensorRT and the default models in the background, shared with other apps that use them

2.2.0

Choose a tag to compare

@github-actions github-actions released this 06 Nov 18:12
921291a
  • TensorRT and the default models are kept once per PC and shared with other apps that use them (yosh's AI add-on, MangaJaNaiCLI); updating moves your install's existing files over, so most of it isn't downloaded again

2.1.0

Choose a tag to compare

@github-actions github-actions released this 05 Nov 04:18
5534275
  • MangaJaNai GUI now runs on Linux (x86_64), on NVIDIA GPUs with the TensorRT backend. It needs the NVIDIA driver and a distro from about 2022 on (glibc 2.35+). Two downloads:
    • MangaJaNaiGUI-v2.1.0-linux-x64.tar.gz: extract it anywhere and run ./install.sh from inside the folder. It adds a menu entry and installs the components for your GPU; the app updates itself from then on, like on Windows
    • MangaJaNaiGUI-x86_64-v2.1.0.AppImage: make it executable and run it (--install adds a menu entry). It includes the TensorRT runtime and the default models; the TensorRT kernels for your GPU download on first launch
  • On Linux, settings, your models and TensorRT engines are kept in ~/.config/MangaJaNaiConverterGui. DirectML and TensorRT-RTX are Windows-only
  • Nothing changes on Windows

2.0.2

Choose a tag to compare

@github-actions github-actions released this 03 Nov 17:50
34669f8
  • Fix tile seams in screentone generated by the MangaJaNai models: tiles now overlap by 128 px and blend in a narrow band at the centre
  • Faster upscaling on NVIDIA (TensorRT and TensorRT-RTX): tiling now runs on the GPU. A 24-page batch with IllustrationJaNai FDAT_M 4x on an RTX 5090 went from 26.2 s to 18.4 s
  • Faster upscaling on AMD and Intel (DirectML): FDAT_M 4x went from 55.3 s to 48.3 s
  • Images with uniform transparency are upscaled once instead of twice
  • Folder upscales now include TIFF and JPEG XL images and 7z/cb7 archives
  • Update to libaji v0.21.0 and MangaJaNaiCLI v0.2.0, the upscaling backend, now in its own repository. Engines built by 2.0.1 are reused