Skip to content

3.12.5

Latest

Choose a tag to compare

@github-actions github-actions released this 20 Nov 17:54
6b700f2
  • AnimeJaNai UltraCompact and SuperUltraCompact models run on the native GPU kernels, 3.5-5.5x faster than on TensorRT
  • HAT, DAT and FDAT models are faster on the native GPU kernels on RTX 30/40/50 cards: 37-49% less time per frame for HAT-L, 22-32% for DAT-2, and TFDAT is 11-16% faster
  • Custom SPAN, Compact, ESRGAN, HAT and DAT models run on the native GPU kernels from any common ONNX export
  • Fix updating from 3.12.3 in the app failing with "Could not find file ... aji_vk.dll"

Downloads

Installer Portable
Windows Setup .exe .7z
Linux AppImage .tar.zst