What is Image-to-image (img2img)?

Image-to-image (img2img) is the AI generation mode where a source image is used as the starting point for a new generation, rather than starting from random noise. It is the core technique behind 'restyle this photo,' 'put this product in a different scene,' and 'turn this sketch into a polished render.' The strength of the source image's influence is controlled by a denoising parameter — low denoising keeps the original layout and details, high denoising lets the model reinvent more freely. img2img is the technical mechanism for product-placement AI photography: the product photo is fed in as the source, the model preserves its silhouette and material properties, and the surrounding scene is generated fresh. It is also the gateway to image-to-video models, where an img2img-style source becomes the first frame and the model animates outward.