Existing pixel-art-specific upscaling methods are going to be better than this, for pixel art. But if they wanted they could probably try to train their neural

Existing pixel-art-specific upscaling methods are going to be better than this, for pixel art. But if they wanted they could probably try to train their neural network for pixel art specifically, similar to https://github.com/nagadomi/waifu2x https://github.com/nagadomi/waifu2x