AnimeGANv3

Why we included this project

AnimeGANv3 is a trained GAN that restyles photos and video frames into a consistent anime look, which sets it apart from the text-to-image models most people reach for first. You feed it real footage and pick a target style, say Hayao or Shinkai, and the released ONNX weights produce output through a small GUI or the command-line scripts without you training anything. The same models convert to TensorFlow Lite, Core ML, and TensorRT, so pushing the stylization into a mobile app or edge device is realistic. Fan artists and short-film makers get a working video-to-anime script out of the box, while engineers can reuse the published inference and conversion code inside their own pipeline. One caveat worth knowing: the weights are distributed under the project's own non-commercial terms, so read those before shipping anything commercial.