How to Upscale Image with Stable Diffusion Models

Have you ever wished you could make your low-resolution images look sharper and more detailed? Or maybe you have some old photos that you want to restore and enhance? Or perhaps you are a fan of AI-generated art and want to create high-quality images with stable diffusion models?

If yes, this article is for you. In this article, we will show you how to upscale image with stable diffusion models, a novel technique that uses AI to improve the quality and resolution of images. We will explain what stable diffusion models are, how to install and run stable diffusion webUI, how to upscale image, and tips for better results.

What are Stable Diffusion Models?

Stable Diffusion Models are a type of machine learning models that can generate realistic and high-quality images from text descriptions. They use a technique called diffusion, which involves adding noise to an input image and then gradually reducing the noise over time to produce a final image.

Stable Diffusion Models are based on the work of Ho et al. (2021), who introduced a new way of training diffusion models that is more stable and efficient than previous methods. Stable Diffusion Models have been used for various tasks, such as creating artworks, cartoons, video games, and more.

Why Use Stable Diffusion Models for Image Upscaling?

Image upscaling is the task of increasing the resolution and quality of an image without losing its original features and details. Image upscaling is useful for many applications, such as restoring old photos, enhancing digital art, improving video quality, etc.

However, image upscaling is also challenging because it requires adding new information that is not present in the original image. Traditional methods for image upscaling rely on interpolation techniques that simply enlarge the pixels or use predefined filters to fill in the gaps.

See also  Unhinged AI – Madness of Artificial Intelligence

These methods often produce blurry or distorted images that lack realism and diversity. Stable diffusion models offer a new way to upscale images using artificial intelligence. Instead of interpolating or applying rules, stable diffusion models generate new information by learning from a large dataset of high-resolution images.

They can create realistic and diverse images that preserve the original features and details while adding new ones. They can also handle different types of images and styles without requiring separate models or training data. Moreover, they can be customized and fine-tuned by adjusting the parameters or the input of the model.

How to Install and Run Stable Diffusion WebUI?

Install Python:

  • First, you need to install Python 3.10.6. You can download “Windows Installer (64-Bit)” from the official Python website.
  • Run the downloaded Python installer and follow the prompts. If you already have Python installed, select “Upgrade.” Otherwise, follow the recommended installation settings.

Install Git and Download the GitHub Repo:

  • Install Git for Windows by downloading the 64-bit Git executable and running it with the recommended settings.
  • Next, you’ll need to download the Stable Diffusion GUI files from GitHub.
  • For AUTOMATIC1111’s WebUI: Go to the GitHub repository and click the green “Code” button, then select “Download ZIP.”
  • For ComfyUI: Scroll down to the “Installing” section in the GitHub repository and click the “Direct Link to Download.”

Extract the Downloaded Files:

  • Open the downloaded ZIP archive using File Explorer or your preferred archiving program.
  • Extract the contents to any location you prefer. This is where you’ll run Stable Diffusion. The example used “C:” but you can choose a different directory.

Download Checkpoints:

  • You’ll need specific checkpoints for Stable Diffusion to work correctly.
  • For AUTOMATIC1111’s WebUI, it may fetch version 1.5 checkpoints automatically. However, for SDXL checkpoints, you’ll need to download them manually.
  • For ComfyUI, you’ll need to download checkpoints manually.
  • Place these checkpoints in the respective folders:
    • AUTOMATIC1111’s WebUI: “C:stable-diffusion-webuimodelsStable-diffusion”
    • ComfyUI: “C:ComfyUI_windows_portableComfyUImodelscheckpoints”

Add Additional Models (Optional):

  • You have the option to add extra models like ESRGAN, Loras, etc., which enhance upscaling quality or provide better results for specific image types.
  • Both ComfyUI and AUTOMATIC1111’s WebUI create folders for these additional models. Simply drag and drop the model files into the appropriate folders.
See also  Best AI Hairstyle Online Free Apps for Man and Woman

Run the GUI:

  • To use AUTOMATIC1111’s WebUI, open the main folder and double-click “webui-user.bat.”
  • For ComfyUI, open the ComfyUI folder and click “run_nvidia_gpu.bat.”

Wait for Completion:

  • Once you’ve launched the GUI, it will start processing images. You’ll see a console display with a local URL, typically “” for AUTOMATIC1111’s WebUI.
  • For ComfyUI, it runs on the same IP address but on port 8188.
  • That’s it! You’ve successfully installed and run Stable Diffusion with a graphical user interface. You can access the interface through the provided local URL to process your images.

Discover more on our blog, where we share tips and tutorials about “3 Ways to Use Stable Diffusion AI Art Generator for Free” Whether you are a beginner or experienced, we have got you covered with valuable insights on the “3 Ways to Use Stable Diffusion AI Art Generator for Free”.

How to Upscale Image with Stable Diffusion WebUI?

Step 1: Upload Your Image

  • Open Stable Diffusion WebUI and navigate to the “Extras” tab, where you’ll find the upscaling tools.
  • If you’ve just created an image you want to upscale, simply click “Send to Extras,” and it will take you to the upscaling section with your image ready.
  • Alternatively, you can drag and drop your image into the provided upload field.
  • If you have multiple images to upscale simultaneously, switch to the “Batch Process” tab. However, note that batch processing may not always work as expected.

Step 2: Choose the Desired Size

  • Use the “Resize” slider to determine the size of the output image. By default, you can use the “Scale By” option, which multiplies the current resolution by a specified factor.
  • For example, if you set it to 2 and your input image is 512×512, it will be upscale to 1024×1024.
  • Alternatively, switch to the “Scale To” option and input a specific resolution. Pay attention to the image’s aspect ratio and uncheck the “Crop to Fit” option if you don’t want any edges cut off.
  • Keep in mind common aspect ratios: 9:16 for phones, 4:3 for tablets, and 16:9 for computers (with ultrawide monitors going up to 21:9).
See also  7 Best AI Product Photography Tools and Services in 2023

Step 3: Choose an Upscaling Algorithm

  • This is where you might feel overwhelmed because there are various upscaling algorithms with cryptic names.
  • Your choice depends on the type of image you’re upscaling, like a photo, painting, anime art, or another “cartoon” style artwork.
  • Different algorithms also have different processing speeds, so consider your time constraints.
  • While experimenting with different algorithms is the best way to find the ideal one for your image, here are some basic recommendations:
  • For Photos: ESRGAN_4x, For Paintings: R-ESRGAN 4x+, For Anime: R-ESRGAN 4x+ Anime6B.

Step 4: Upscale Your Image

  • Once you’ve configured your settings, click the “Generate” button to begin the upscaling process.
  • Keep in mind that the first time you use a particular algorithm, Stable Diffusion will need to download the necessary models, so it might take longer initially, depending on your internet connection.
  • After upscaling is complete, you’ll find the output images in the “extras-images” subdirectory within your “outputs” folder.
  • You’ve successfully upscaled your image using Stable Diffusion WebUI. It’s a matter of choosing the settings that suit your image type and clicking the “Generate” button to obtain the upscaled result.

Tips for Better Upscaling Results

Use High-Quality Input Images

The quality of the input image affects the quality of the output image. If the input image is too small, blurry, noisy, or distorted, the upscaling process may amplify these defects and produce artifacts or blur. Therefore, it is recommended to use high-quality input images that have a reasonable resolution, sharpness, contrast, and color.

Experiment with Different Upscalers

Different upscalers may work better for different types of images or styles. For example, Waifu2x and Anime4K are designed for anime-style or cartoons, while Real-ESRGAN and CUnet are more general and can handle natural and stylized images. You can try different upscalers and compare their results to see which one suits your needs and preferences.

Adjust the Settings and Parameters

Each upscaler has its own settings and parameters that control how it upscales images. For example, you can adjust the noise level, the model size, the number of steps, the interpolation method, etc. You can tweak these settings and parameters to fine-tune the upscaling process and achieve better results.

Frequently Asked Questions


In this article, we have discussed how to upscale image with Stable Diffusion Models, a type of machine learning models that can generate realistic and high-quality images from text descriptions. We have explained the technique of diffusion, which involves adding noise to an input image and then gradually reducing the noise.

We have also introduced some of the advantages of Upscale Image with Stable Diffusion, such as preserving the details and quality of the original image, learning the underlying patterns and structures of the image, and generating realistic textures and features.