Conditional diffusion models offer new tradeoff points between distortion and perception in image compression, enhancing visual quality at low bitrates.
