What does denoising strength actually control?

It sets how much noise is added to your source image before the model re-denoises it. At 0.0 the output is identical to the input; at 1.0 the image is fully noised and the result ignores the original almost entirely.

Why does my img2img output look unchanged?

Your denoising strength is too low — typically below 0.3. The model only repaints the noise it added, so a low value preserves nearly all the original pixels. Raise the strength to give the model room to change the image.

What is a safe default to start with?

0.5 is a sensible middle ground for most edits. It keeps the overall composition and pose while letting the model restyle colours, textures and details. Adjust from there based on how much change you want.

Does denoising strength affect generation speed?

Indirectly. The number of actual sampling steps run is roughly steps × denoising strength, so a lower strength finishes faster. Some UIs show this as the "effective steps" for the img2img pass.

img2img Denoising Strength Advisor

img2img denoising strength, explained

Denoising strength is the single most important slider in Stable Diffusion’s img2img tab, and the most misunderstood. It is a value from 0.0 to 1.0 that decides how far the model is allowed to drift from your source image. Pick the wrong value and you either get an output identical to the input or one that has nothing to do with it. This advisor maps your transformation goal and source image type to a recommended starting range so you stop guessing.

How denoising strength works

When you run img2img, the pipeline does not start from pure random noise the way text2img does. Instead it takes your source image, adds a controlled amount of noise to it, and then denoises back toward a clean image guided by your prompt. Denoising strength is the size of that noise injection:

0.0 — no noise added, output equals input.
~0.3 — light repaint; pose, layout and colours are preserved.
~0.5 — balanced; composition kept, style and details reworked.
~0.75 — heavy reinterpretation; only rough shapes survive.
1.0 — effectively text2img; the source barely matters.

Because the sampler only runs across the noised portion, the effective steps are about steps × denoising_strength. Low strength is therefore both subtler and faster.

Tips and worked examples

Sketch to render: start around 0.6–0.75. A loose drawing has little detail to preserve, so you want the model to invent texture and lighting.
Photo restyle (e.g. anime filter): 0.45–0.6 keeps the subject recognisable while swapping the art style.
Fixing a hand or small artifact via inpainting: 0.3–0.5 on the masked region; higher and the fix won’t blend with the surrounding pixels.
Iterate in small steps. Change strength by 0.05 at a time and keep the seed fixed so you can see the effect of the slider alone.