ComfyUI Image to Image: Denoise Settings and Common Mistakes
Learn how ComfyUI img2img works, how to tune denoise, and how to avoid common reference-image mistakes.
What Image to Image does
Image to Image, often called img2img, starts from an existing image instead of pure noise.
In ComfyUI, it is one of the fastest ways to restyle an image, preserve composition, or push an existing render in a new direction.
The most important setting: denoise
The denoise value controls how much the model is allowed to change the input image.
0.0to0.3: very small changes0.3to0.6: balanced changes0.6to0.8: stronger transformation0.8to1.0: close to starting over
Practical rule
If you want the result to still respect the reference image, keep denoise below 0.9.
Common mistakes
The result looks too similar to the original
- Increase
denoise - Use a clearer prompt
- Make sure the checkpoint matches your goal
The result ignores the reference image
- Lower
denoise - Check whether your prompt is overpowering the input
- Use a better source image with stronger structure
The reference image looks distorted
- Make sure the image dimensions are divisible by
8 - Avoid very small images (
<256px) - Check whether the VAE and checkpoint are compatible
Recovery tips
If img2img starts failing after plugin installs, node updates, or environment drift, the problem may not be the workflow itself. It is often the environment.
Use these guides next:
- How to Fix ComfyUI Plugin Import Failed Errors
- ComfyUI Startup Failed? How to Diagnose and Recover Faster
- ComfyUI Dependency Conflicts
Need a faster path?
If your ComfyUI environment is already unstable, start with Wonderful Launcher. If the environment is still blocked, use expert help.
Wonderful Launcher ドキュメント