Upload an image to see how MAGVITv2 vector quantizes and reconstructs it. The model downsamples the image by a factor of 16 into discrete tokens.