Image-to-image
Image-to-image (img2img) is the AI workflow that transforms an existing picture according to your prompt, keeping some of the original structure instead of generating from scratch.
Image-to-image - usually shortened to img2img - feeds an existing picture into the model alongside your prompt. Instead of starting from pure noise, the model starts from a partially noised version of your image, so the output keeps the original's composition, pose or layout while taking on the new style or content you describe.
How it works
The source image is encoded into latent space, then a controlled amount of noise is added. The model runs denoising from that point, guided by your prompt. The key dial is denoise strength (sometimes called denoising strength or "transformation amount"):
- Low strength (~0.2-0.4) keeps the original almost intact - good for subtle restyling, color changes or cleanups.
- Medium strength (~0.5-0.7) keeps the overall composition but reinvents textures, lighting and detail.
- High strength (~0.8-1.0) barely respects the source and behaves almost like a fresh text-to-image generation.
Why it matters
Img2img is how you iterate. Generate a rough draft (or upload a sketch or photo), then nudge it toward the look you want without losing the parts you already like. It is the backbone of restyling, sketch-to-render and consistent variations.
Upload a quick phone photo of a room, then restyle it with a medium denoise strength so the layout stays but the mood changes.
Example prompt
Same room interior, restyled as a moody mid-century lounge at dusk, warm tungsten lighting, cinematic - denoise strength 0.6Try it in the generator
Put image-to-image to work right now - free daily generations, commercial license included.
Frequently asked questions
Is image-to-image the same as img2img?
Yes. "img2img" is simply the shorthand spelling of image-to-image. They refer to the same workflow: transforming an input image with a prompt instead of generating from noise.
What does denoise strength do in img2img?
It sets how much of the original image is kept. Low values stay close to the source; high values let the model deviate freely and behave almost like a fresh text-to-image generation.
Related terms
- img2imgimg2img is shorthand for "image-to-image" - the AI workflow that transforms an existing picture using a prompt, keeping some of the original structure instead of generating from scratch.
- Text-to-imageText-to-image is the AI workflow where you type a written prompt and the model generates a brand-new image from it - no source image required.
- DenoisingDenoising is the core operation of a diffusion model: at each step it predicts and removes a little noise, gradually turning a random field into a clear image.
- InpaintingInpainting regenerates only a selected (masked) region of an image while leaving the rest untouched - useful for removing objects, fixing details or replacing part of a scene.