What is Image-to-image?
Also known as: img2img
A diffusion model mode that takes an input image and a prompt, producing an output that combines the input's structure with the prompt's style. The opposite of text-to-image.
In detail
Image-to-image (img2img) starts the diffusion process from a noised version of an input image instead of pure noise. A 'denoising strength' parameter (0.0-1.0) controls how far from the input the output drifts: 0.0 returns the input unchanged, 1.0 essentially ignores it (equivalent to text-to-image with the input as the seed). Strength 0.3-0.5 preserves overall composition while changing style; 0.6-0.8 keeps rough structure but allows style transformation; 0.9+ uses input only as an aesthetic hint. Textile applications: applying watercolor style to a vector illustration (low strength), transforming a photograph into a textile pattern (medium), extending a small motif into a full design (medium-high). Image-to-image generation is the dominant technique for refining or restyling existing artwork — turning a hand-sketched motif into a watercolor render, converting a photograph into a stylized pattern, or adapting an existing print to a new color palette. The denoising strength parameter controls how much of the source survives: 0.3 preserves most of the original, 0.8 mostly disregards it.
Example
A photograph of fresh peonies + prompt 'soft watercolor textile pattern, pastel palette, ditsy layout' at strength 0.45 produces a watercolor floral pattern that retains the original peony arrangement but has been re-rendered as a textile-friendly watercolor. The eye recognizes the source peonies but the result is design-ready.