Long time no see, Anime4KCPP v3.2 is ready.
Anime4KCPP v3.2 optimizes performance and introduces more models.
First are ACNet and ARNet, which have been redesigned and retrained. The previous ACNet model is now referred to as ACNet-Legacy. ACNet and ARNet offer models of various sizes, which can be selected based on your hardware configuration. Each size provides 4 variants:
no suffix: trained to be neutral.hdn: trained to mildly denoise.box: trained to be neutral through box degradation, which is better for line restoration, but may appear slightly blurry.box-hdn: trained to mildly denoise based on the box variant.
If you are interested in the training of ACNet and ARNet, or the scripts for exporting them to GLSL, please check out the training code.
We have also integrated Artoriuz's ArtCNN series, igv's FSRCNNX series, and nessotrin's FSRCNNX-DistortPlus series. Thanks for their wonderful work.
Full model list is available in wiki. You can try these models and test their effects directly in your browser via the WASM-driven Anime4KCPP Playground.
In terms of performance, whether using CPU, OpenCL, or CUDA processing, v3.2 is faster than v3.0. With smaller models such as ACNet_F8B4, Anime4KCPP can achieve real‑time processing even on low‑end hardware.
Besides, we have added HDR support in the DirectShow filter, as well as support for YUV422 and YUV444 video formats. In the video module, we support more formats, including 12‑bit video and grayscale video. Various other detail improvements and optimizations are also included.
Please ensure the latest Visual C++ Redistributable is installed prior to using the precompiled MSVC binaries.
Check wiki for more information.