Stable Diffusion Denoising Strength: What Is It and How to Use It

Stable Diffusion, a revolutionary text-to-image AI model, has captivated creators and enthusiasts with its impressive ability to generate stunningly realistic and imaginative visuals. At the heart of this powerful tool lies a crucial parameter known as Denoising Strength, which plays a pivotal role in shaping the final output of your generated images.

What Is Stable Diffusion Denoising Strength

Denoising Strength: Determines how little respect the algorithm should have for image's content. At 0, nothing will change, and at 1 you'll get an unrelated image. With values below 1.0, processing will take less steps than the Sampling Steps slider specifies.

Stable Diffusion initiates the process of image-to-image generation process by adding noise to the input image based on the seed. The amount of noise added is determined by the Denoising Strength value, which ranges from a minimum of 0 to a maximum of 1.

A lower denoising strength means the output image will be more similar to the input image, with only minor changes. Thus, a Denoising Strength of 0 will add no noise at all, so the output will be identical to the original input.

A higher denoising strength allows the model to create a more divergent, new image that looks quite different from the original input. At a denoising strength of 1, the output will be a completely new image.

Thus, you can understand Denoising Strength as the factor that controls the balance between preserving the input image and generating a completely new output image.

FYI: Stable Diffusion does an amazing job of turning your text prompts into creative artwork. However, its default output resolution is 512p or 768p. While you can set higher resolutions, this often leads to increased strain on your computer, causing freezes or out-of-memory errors. The recommended workflow is to create low-resolution images with Stable Diffusion and then upscale them using Aiarty Image Enhancer.

Aiarty Image Enhancer

Upscale and Enhance AI-generated Images with AI

• Increase pixel quality while keeping and generating details
• Fix imperfections like pixelation, graininess, and blurriness
• Upscale image resolution to 4K, 8K, 16K, or even 32K

Free Download Free Download

Free Download

How to Use Denoising Strength in Stable Diffusion

Before diving into denoising strength in Stable Diffusion, consider your image goals. If you're aiming for an output that looks similar to the starting image, you might want to opt for a lower denoising strength. Suppose you need a more unique and diverse outcome, a higher denoising strength is more recommended. You can find the denoising strength setting in the img2img, sketch, and Inpaint on the Stable Diffusion Web UI, near where you upload your image.

How to Use Denoising Strength in Img2img Mode

Step 1. Open img2img. Go to the Generation section and choose img2img.

Step 2. Add the starting image to the image canvas by dragging and dropping.

Step 3. Go to the Prompt section. Enter the text prompt that clearly describes the desired changes or additions you want in the final work. If necessary, go to the Negative Prompt section and enter the text prompt that specifies any elements or styles you want to avoid in the final piece. You can also click the Interrogate CLIP button to guess the prompt of the input image.

Step 4. Adjust the Denoising Strength value. Click Generate to generate a new image.

Remember, there's no universal denoising strength setting – it's about finding the right balance that fits your preferences and needs.

How to Use Denoising Strength in Inpaint Mode

The denoising strength setting in Stable Diffusion's inpainting mode (Inpaint Sketch and Inpaint Upload) works similarly to its usage in the image-to-image workflow. Adjusting this parameter allows you to control the balance between preserving the original image and introducing more changes. Below is a demonstration of how to use Denoising Strength in the Upload Inpaint mode.

Step 1. Open the img2img tab on Stable Diffusion. Hit Inpaint Upload and select the original image for uploading to Stable Diffusion.

Step 2. Navigate to the Prompt section and enter the text prompt that details the changes or additions you wish to see in the final work. If there are specific elements you want to avoid, go to the Negative Prompt section to list these exclusions. Asides, you can click the Interrogate CLIP icone to generate the prompt automatically from your input.

Step 3. Select the area that you want to make changes by using the brush. Check the option box of Inpaint masked for recreating the masked area.

Step 4. Adjust the Denoising Strength value as desired. A higher value will introduce more changes to the masked area, while a lower value will preserve more of the original image.

It's important to note that you should be cautious when setting the denoising strength too high in inpainting. Excessively high values can result in an incoherent generation that doesn't blend well with the rest of the image.

Optional Reading about Denoising Strength

When working with generative models like Stable Diffusion, the denoising strength is a crucial parameter that can significantly impact the quality and characteristics of the output images. The denoising process is a key component of Stable Diffusion's workflow, and understanding how it works can help you leverage this capability to improve the denoising strength for your specific need.

1. How Does SD Image to Image Generation Work

Stable Diffusion's workflow is mainly divided into 2 stages:

Forward Diffusion: In the forward diffusion stage, the model gradually transforms a clear image into a featureless noise image. This process is akin to gradually adding noise to an image until its original content becomes unrecognizable.

Reverse Diffusion: However, Stable Diffusion's goal is not just to transform images into noise. Its true goal is to recover the original clear image from the noisy image through reverse diffusion. This requires training a neural network model to predict the added noise. This model is called the Noise Predictor, and it is a U-Net model.

2. Why the Stable Diffusion Denoising Strength Matters

Now that you have a rough idea about the forward and reverse diffusion processes in Stable Diffusion, you can more easily understand why the denoising strength parameter is so crucial when doing image-to-image generation.

The key steps in Stable Diffusion's image-to-image generation workflow are:

Step 1. Encode the Input Image: The input image is encoded into the latent space using an encoder, typically part of a Variational Autoencoder (VAE). This converts the image into a compressed, latent representation.

Step 2. Add Noise to the Latent Space: Noise is added to the latent space representation of the input image. The amount of noise is controlled by a denoising strength parameter. If the denoising strength is set to 0, no noise is added; if it is set to 1, the maximum amount of noise is added, transforming the latent representation into a completely random tensor.

Step 3. Predict and Subtract Noise: The U-Net, a neural network architecture, takes the noisy latent image and the text instruction as inputs. It predicts the noise present in the latent image. This predicted noise is subtracted from the latent space image. This step refines the image by reducing the noise.

Step 4. Iterative Denoising: Steps 3 and 4 are repeated for a specified number of sampling steps (e.g., 20 times). This iterative process gradually refines the image, aligning it more closely with the text instruction at each step.

Step 5. Decode to Image Space: The final refined latent space image is decoded back into pixel space using the VAE decoder, producing the final output image.

So now you know why denoising strength is so important for image-to-image generation in Stable Diffusion. It sets an initial latent image with a mix of noise and input image. When the denoising strength is 1, it's equivalent to text-to-image, as the initial latent image is completely random noise.