Image-to-Image (img2img)
Image-to-image, or img2img, generates a new image by transforming an existing one according to a text prompt instead of starting from scratch.
June 16, 2026
Image-to-image (img2img) is a generation method that takes an existing image plus a text prompt as its starting point and produces a new image, transforming the original while preserving the structure you choose to keep.
How it works
Instead of beginning from random noise, img2img seeds the process with your source image. The model adds a controlled amount of noise to that image and then denoises it back, guided by your prompt. Because it starts partway through the diffusion process rather than at the very beginning, the broad shapes and composition of the original tend to survive.
A strength (or denoising) value sets how much of the original is overwritten. A low value nudges the image — recoloring, light restyling — while keeping it recognizable. A high value gives the prompt more freedom to reinterpret the scene, so the output can diverge significantly from the source. This makes img2img a flexible bridge between a reference you already have and the result you are imagining.
Why it matters
Img2img gives you a way to direct generation with a concrete visual starting point, not just words. That is powerful for restyling photos, finishing rough sketches, iterating on a composition you like, or producing consistent variations. It reduces the guesswork of pure text prompting because the layout is already established, and it lets non-artists transform images quickly.
In eaxy
eaxy is built around generation, with style packs and motion at the center. Where eaxy supports starting from a reference, the same idea applies: give it an image and a direction, choose a style, and let it generate a fresh take you can then animate with Kling 3. The focus stays on creating new visuals fast rather than pixel-level manual editing.
Related terms
Frequently asked questions
What is img2img?+
Img2img is short for image-to-image. You provide a starting image and a prompt, and the model produces a new image that keeps some of the original's structure while applying the changes you describe.
How is img2img different from text-to-image?+
Text-to-image starts from pure noise with only your words to guide it. Img2img starts from your existing image, so the output stays closer to the original composition, pose, or layout.
What controls how much img2img changes the image?+
A strength or denoising setting decides how far the result departs from the source. Low values make subtle edits; high values let the prompt reshape the image more freely.
What is img2img used for?+
Common uses include restyling a photo, turning a rough sketch into finished art, exploring variations of a scene, and adjusting the look of an existing render while preserving its layout.
Make it with eaxy
Describe anything and generate stunning images in seconds — then bring them to motion with Kling 3.