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