✨小透明・宸✨

稍等一段时间，这里的模型有点多😂 之后有空的时候我试着转换一下。

https://github.com/upscayl/upscayl/wiki/Model-Conversion-Guide 是这个教程吗？ 在 chaiNNer 的源代码里翻到了这个 https://github.com/chaiNNer-org/chaiNNer/blob/main/backend/src/nodes/impl/onnx/onnx_to_ncnn.py ，应该就是实现转换的代码，不过有点复杂，不知道能不能分离出来以及会不会比现在我这里的那个脚本更好用。

I haven't paid much attention to new models and still using `realesrgan-x4plus-anime`, I think it's quality is enough. You can also give Real-CUGAN a try.

They all comes with Real-ESRGAN/CUGAN's binary. https://github.com/xinntao/Real-ESRGAN/releases https://github.com/nihui/realcugan-ncnn-vulkan/releases You can find more ESRGAN model from [OpenModelDB](https://openmodeldb.info/?t=arch%3Aesrgan). They are in pth format and you need to convert to onnx format to...

Both Real-ESRGAN-ncnn-vulkan and upscayl-ncnn only support 2-4x for now, maybe I could have a try of adding 8x by some copy-and-pasting in `realesrgan.{h,cpp}` and then make a custom build by...

Can you attach some converted 8x (or even 16x?) NCNN model (`bin` and `param` file) for me to do some testing? --- Take some note about upscayl-ncnn's code related to...