The technology can upscale 2D images and even anime.
Take a look at Real-ESRGAN – an open-source AI image upscaler trained with pure synthetic data. It extends ESRGAN (Enhanced Super-Resolution Generative Adversarial Networks) to a practical restoration application to improve low-resolution images by upsacaling them and reducing noise.
Real-ESRGAN supports 16-bit images, gray images, tile options, and images with alpha channels. With time, the tool has been improved with several AI models, like Tencent's GFPGAN to support face enhancement. It can also upscale anime images and even videos pretty well: