Image-to-image (img2img)
Image-to-image starts a diffusion generation from an existing image instead of pure noise, restyling or refining it.
Image-to-image (img2img) seeds a diffusion run with your input image rather than random noise, so the output inherits the source's composition, pose, and layout while the prompt re-renders its style, materials, and detail. How far it strays is set by denoising strength. Compared with inpainting (which edits a masked region) and outpainting (which extends the frame), img2img transforms the whole image at once. It is the workhorse for turning sketches into renders, photos into illustrations, low-res drafts into polished frames, and for iterating on a near-miss generation without losing what already works.
When to use image-to-image (img2img)
- Restyling a whole image while keeping its composition.
- Turning a rough sketch, 3D blockout, or photo into a finished render.
- Refining a generation you almost like (low denoising strength).
When not to use image-to-image (img2img)
- Editing one small area only — use inpainting for surgical control.
- Starting from nothing — that is plain text-to-image.
Example
Input: img2img from a pencil sketch, denoising 0.6, prompt: "oil painting, dramatic side light" Output: A painterly render that follows the sketch's lines but adds color, light, and texture.
Common mistakes
- Confusing img2img with inpainting — img2img changes the entire frame.
- Cranking denoising near 1.0 and losing the source entirely.
- Assuming low strength fixes a weak prompt; structure is kept, quality still tracks the prompt.
FAQ
What is image-to-image (img2img)?
Image-to-image starts a diffusion generation from an existing image instead of pure noise, restyling or refining it.
When should I use image-to-image (img2img)?
Restyling a whole image while keeping its composition. Turning a rough sketch, 3D blockout, or photo into a finished render. Refining a generation you almost like (low denoising strength).
What are the most common mistakes with image-to-image (img2img)?
Confusing img2img with inpainting — img2img changes the entire frame. Cranking denoising near 1.0 and losing the source entirely. Assuming low strength fixes a weak prompt; structure is kept, quality still tracks the prompt.
Related terms
- Inpainting — Inpainting regenerates a masked region inside an image while keeping the rest untouched, for targeted edits.
- Denoising strength — Denoising strength controls how much an image-to-image run reshapes the source image, from 0 (untouched) to 1 (source ignored).
- CFG scale (classifier-free guidance) — CFG scale controls how strongly a diffusion image model follows its text prompt — higher values stick closer to the prompt, lower values explore more.
- Diffusion model — A diffusion model is a generative neural network that creates images, video, or audio by iteratively denoising random noise toward a learned target distribution.
- Seed — A seed is an integer that initializes the random number generator inside an image, video, or audio model, making generation reproducible.
Sources
Last updated: 2026-06-02. Raw markdown: https://promtable.com/glossary/img2img.md.