What is Image-to-Image (img2img)?
TL;DR
An AI method that takes an existing image as input and generates a new one with changed style or content.
Image-to-Image (img2img): Definition & Explanation
Image-to-Image (img2img) is an AI method that takes an existing image — not just text — as input and transforms its style, texture, or content while preserving its features to generate a new image. In diffusion models, the 'denoising strength' controls how much of the original is kept when noise is added and the image is regenerated: lower values stay closer to the original, higher values change it more boldly. It has wide applications, such as converting photos to anime or figurine style, cleaning up rough sketches, and swapping clothing in virtual try-on. It is the counterpart to Text-to-Image and is valued whenever you want to preserve composition or subject.