Upscayl - Enlarge and sharpen photos with AI

With the Upscayl app, it should be possible to scale up small photos on Mac, Windows PC and Linux machines while maintaining a certain sharpness. Blurred images are to be made sharper with Upscayl by means of a few clicks, so that they look more detailed. I've tested the free app and find that it can definitely help here and there. However, I also encountered a few weaknesses and bugs. That's why I made Upscayl from the app recommendations published yesterday removed and the observations made during testing are explained in more detail below.

TL; DR: App website / Project on GitHub

Features of the Upscayl app at a glance

  • A software primarily developed for Linux, which is also available for macOS and Windows
  • Can upscale images and photos to show more detail than just zooming in
  • Can add more detail and sharper edges to blurred images, making them look sharper
  • Multiple files can be processed in one go (batch processing)
  • Various (even self-added) AI models can be used to achieve the desired result

Download and install the app

You can get the free app, which was designed as an open source project, from their official website. There is a reference to the current version (currently v2.5.1) and the fact that there are already over 1 million downloads. Below that are the download buttons for macOS, Linux, and Windows. A dmg file is offered for the Apple Mac, after which you simply mount the Appicon into the Applications folder. 

However, the first time you try to open it, it could happen that the Mac Gatekeeper indicating that Upscayl will not open "because the developer cannot be verified". In this case, click on "Cancel", open the system settings, go to the "Privacy & Security" area and click on the "Open anyway" button when the Upscayl blocking message appears. When you try to open it again, click on "Open" again and then you can finally start.

The Upscayl interface: tidy and intuitive

Once opened, the program doesn't overwhelm you with too many options or complicated menus. It is kept simple and offers only two tabs: "Upscayl" for using the core function and selecting the necessary settings. And "Settings" for settings that affect the program and the general conditions of use.

In the Upscayl tab you can select an image file under Step 1. This must have been saved in JPEG, JPG, PNG or WEBP format. In Step 2, an AI model or the associated upscaling method is selected. If you tick the “Double Upscayl” box, the process will be carried out twice and the target file will be enlarged 16 times. In Step 3, a target folder is specified and in Step 4, the process is started.

The "Batch Upscale" slider can also be activated at the top. Once this is done, not just a file can be selected for editing, but an entire folder full of images. Here, too, the photos or graphics must be available in the above-mentioned file formats. Then the steps mentioned can be worked through as usual.

In the Settings tab you can make a few basic settings that also affect the upscaling process described above. First you can set a design for the app. In addition to the light and dark modes, there are many others that sometimes produce wild color combinations. But the following settings are more important: set a permanent target folder (to skip step 3), a GPU Enter the ID (I don't know what the purpose of this is yet), add more AI models, set a target file format (PNG, JPG or WEBP), set the scaling size, etc.

Good performance, but sometimes with "filter" effects

I have a few for my little (and by no means exhaustive) test Stock footage images by Pexels downloaded, saved them in different sizes and then let the Upscayl app edit the smallest ones. I noticed that certain scenes with simple surfaces and structures are processed better than complex patterns. But these are also upgraded to a certain extent. When scaling up a wall of stones, however, recurring hatching became visible under some settings, reminiscent of an art filter from effects apps or social media filters. Nevertheless: You can quickly make photos larger and sharper with the app.

Experimental settings may lead to undesired results

If you look at the different options in the Settings tab, you will find "Image Scale" among other things. By default, this multiplier is set to "4x" for enlarging image files (hence the 16x enlargement described above when running twice). For now, you should leave it there. I tried "3x" as a test and instead of an upscaled image I got an upscaled mosaic whose individual parts were sharper and larger, but the original photo was rather fragmented. So you have to wait for an update for the setting to work properly at some point.

Depending on the size and level of detail, the result will be washed out

I tested a few old digital photos that were quite small and had writing and e.g. B. have shown details of a cobblestone street. When scaling up with Upscayl, the image was generally larger and more attractive, but if you took a closer look at the individual elements, you noticed a "washing out". The writing was not sharper per se, but rather rounded, the cobblestones looked like they had been washed (details were largely softened) and so on. From a distance, however, it looked better than the original. My conclusion from this: upscaling is a compromise and you should not have too high expectations.

Use your own AI models for better results

In my Contribution to Mochi Diffusion and other AI apps for image generation I have already pointed out that there are various AI models for this and similar applications. These represent differently trained AIs that can deal with the source material or with the text prompt in an individual, differently oriented way. In the art AI programs, e.g. B. create different art styles. 

In Upscayl you can also use specially trained models or models downloaded from the Internet in addition to the already integrated AI models. You can easily implement them by clicking on “Select Folder” under “Add Custom Models” in the Settings tab and then selecting the folder in which you saved the models. I haven't tested it myself (yet), but I didn't want to leave it unmentioned. 

By the way: If the option is missing for you, you will find information about the necessary repository here at GitHub.

Conclusion on the short Upscayl test

A small image file or a blurry photo can only be sharpened and detailed to a certain extent. But with Upscayl you can download a free app onto the Apple Mac, Windows PC or Linux computer, which you can use to experiment with improving the image quality. Especially for people who like Photoshop, Affinity Photo, GIMP and Co. are too complicated, this solution, which specializes in one function and can be operated with just a few clicks, could be just the thing. It costs nothing to try. And if you also have experience with Upscayl, then you can leave a comment :)

Did you like the article and did the instructions on the blog help you? Then I would be happy if you the blog via a Steady Membership would support.

2 comments on “Upscayl – Enlarge and sharpen photos thanks to AI”

  1. Hello Johannes, I tried Upscayl after your tip and I have to say that the results surprised me, I didn't believe that it was possible. I work in Graphic Design & Desktop Publishing. Very often the images come from the client with very low resolution, or even from websites. So far I've used PhotoZoom Pro to interpolate the images, from now on I'll only be able to do that with Upscayl.
    Thanks for tip!

    1. Johannes Domke

      Hello Andrej,

      I am glad that the tip helped you and that you are satisfied with the results. I stumbled across this app by accident and found the results in my little test to be quite okay. That's how the post came about. Nice that he has already helped a few people :)

      Best regards
      John

Post a comment

Your e-mail address will not be published. Required fields are marked with * marked

In the Sir Apfelot Blog you will find advice, instructions and reviews on Apple products such as the iPhone, iPad, Apple Watch, AirPods, iMac, Mac Pro, Mac Mini and Mac Studio.