Ben Patterson Read more December 4, 2022

Digital artists and content creators are excited about using deep-learning text-to-image platforms to create out-of-this-world images. Generating accurate images based on keywords has completely changed the game of digital art. However, some who use the text-to-image platform Stable Diffusion have a common complaint. Although the images created may be exactly what they’re looking for, image size has been a huge disappointment.

Creating the perfect piece of digital art using Stable Diffusion is fantastic, but what good is it if the image size is far too small to use? Fortunately, there are ways to upscale an image. In this article, we’ll discuss how to upscale images from Stable Diffusion.

By default, Stable Diffusion’s default image size is 512 x 512 pixels. This native resolution is considered small in today’s digital world and presents challenges to those who need to use files created with Stable Diffusion in a much larger format. The model was trained on datasets of 512 x 512 resolution images and therefore its output is in the same format. But most social media platforms require a resolution of 1080 x 1080 for acceptable viewing, which truly highlights how important an image’s resolution is.

Stable Diffusion can create the perfect image that fits all of the designer’s desires. However, if the file size is so small as to render it unusable, it does present a major dilemma. Simply dropping the image into an app and increasing its size will result in severe image degradation, most likely worse than the original file. Images with such low resolution won’t print well and also cannot be placed into Photoshop to be edited to a designer’s satisfaction.

Fortunately, there are ways to upscale a low-resolution image created with Stable Diffusion. Some users have been quite creative in developing techniques to accomplish this. One difficult and time-consuming method is to split up an image into smaller 512 x 512 sections and then stitch them back together. Others use custom-made algorithms to regenerate AI images into higher-resolution formats.

There are numerous online tools you can use to upscale your images. However, most of these tools ask you to create an online account by using your email address or to pay for services. If you’re looking for a free tool you can use countless times without a subscription, TinyWow is an excellent choice.

You can chaiNNer to upscale Stable Diffusion images. It’s a flowchart/node-based image processing GUI (graphical user interface) that helps to chain image processing tasks. Its strong point is upscaling images. You’ll have full control over your processing pipeline by connecting nodes. This makes it much easier to perform incredibly complex tasks by letting chaiNNer do the work for you.

It works with Windows, macOS, and Linux. If you’re new to working with GUIs, chaiNNer might seem difficult at first. Fortunately, it’s not a difficult process to use chaiNNer to upscale images. By dragging and dropping specific nodes, you can set up a flowchart of processes to do all the heavy lifting. Here’s how to start using chaiNNer:

It should be noted that any other images located in the directory that you select will also be processed. If you only want to upscale one image, you’ll first need to remove any others located in the same directory. However, since the upscaling process takes a lot of time, it would serve the user to have any images that require upscaling to be in the same folder so chaiNNer can upscale them at the same time. Once you have the appropriate image or images in the same directory, you can move on to the next steps.

Next, you’ll need to check how much this model will upscale the image you’ve chosen. The upscale sizes are preset, so this is why you’ll want to ensure that it’s large enough before running the process. If you need the image larger than the preset size, you can run the process twice to double its size. Here’s how to do it:

It could be beneficial to have a specific folder labeled “Upscaled Images” so that you can keep the original image in one folder and the upscaled one in another. Once everything is to your liking, hit the green arrow on the top of the screen to begin the process. The lines you drew will begin to animate and will continue to do so until the process has been completed.

The upscaling process is very time-consuming and can take hours to complete. Once the process has been completed, however, you can check your images by opening them from within the Upscaled Images folder or whichever folder you designated for the final image. Please note that the upscaled images will be significantly larger than their originals. It’s crucial to ensure that you have adequate disk space before beginning the upscale process.

Using chaiNNer may seem complex, but once you get comfortable with its flowchart functions, it’s actually quite simple to use. Using nodes and attaching them with chains (lines) shows exactly how the entire process will function and is great for visual learners.

Other than the 512 x 512 default image size, there are other notable issues that users of Stable Diffusion report.

The rendering of faces can sometimes pose problems, especially when the desired result is photorealistic. For creators who want an anime or surrealist face, this generally isn’t a problem. However, if you’re looking for an authentic and natural look, at times Stable Diffusion can fail. This is because there is no method to have an AI-generated image focus only on the face. However, it’s possible to zoom in and re-render the face for better results.

Another issue worth noting is the proper rendering of human limbs. Again, this is only of concern when you want the desired image to be photorealistic. Sometimes limbs are rendered improperly or in unnatural positions. Users have reported images being generated with extra limbs, and sometimes extra fingers on hands.

These issues should become less frequent as the researchers at Stable Diffusion add more datasets and finetune their algorithms.

Stable Diffusion is an interesting text-to-image platform. Although it produces small files with low resolution, they can be upscaled. Although it might be time-consuming, the process is necessary if you want to further edit an image or ready it for printing. Using chaiNNer, images can be made significantly larger without losing image quality.

Have you tried to upscale an image created with Stable Diffusion? Did you use chaiNNer? Let us know in the comments section below.

It’s now 768×768 with Stable Diffusion 2, which was released a few days after this article was released.

Your email address will not be published. Required fields are marked *

Save my name, email, and website in this browser for the next time I comment.

Disclaimer: Some pages on this site may include an affiliate link. This does not effect our editorial in any way.

Please enable JavaScript to submit this form.

Contact Us | Privacy Policy | TOS | All Rights Reserved