I have always been searching for good image
upscalers that give the best result in the easiest way possible, and preferably
free. Today, we are going to go through this open-sourced AI image upscaler,
find the best settings for it to work, see how you can install different custom
models to fine-tune it, and answer the most important question: can it replace my
existing upscaling software? Let's find out. In my day-to-day usage, I
mainly use two image upscalers: Adobe Lightroom and ChaiNNer. Adobe Lightroom
has a very smooth and user-friendly interface, but it is expensive for most of the people that
are just starting out. On the other hand, Chainer is very versatile but a little bit complicated for
most people due to its node-based structure. So, I needed something that is either cheap
or free, I can easily use on the fly, can be installed and used offline, and works
as good as, if not better than, these two. And that's when I stumbled upon Upscale.
You can check out their GitHub repo, or you can directly go to upscayl.org and
download the installation file. You do not need to sign up to download this software.
There are Linux, macOS, and Windows versions available for this. I just need to mention
one thing: if you are downloading for macOS, it is preferable to download the DMG
format directly because downloading from the Mac Web App Store costs around $13.
Just download this software and install it. Once installed, it shows a pretty basic and
intuitive workflow. Click this button to select the image. It already comes pre-equipped with six
different upscaler models, and for most parts, those are already enough. They have also
mentioned which model is best for which type of images. Generally, the upscale
models provided are 4X upscale models, but you can also select 16X upscale by checking
this box here. It upscales your image twice, thus making it 4X into 4X (16X upscale). Not only that, you can even upload an entire
folder and get all the images upscaled at once. Just select "Batch Upscale" and select
the folder consisting of the images that you want to upscale. That's it. Let us check out
the settings page for more minute controls. You can change the theme from here. There
is an option to save the image as PNG, JPG, or WebP. You can select whether you want to
post-process the images after they have been upscaled. You can see 4X upscale is selected as
the base model here, but you can change it to even 2X or 3X. And the most interesting part is right
here: add custom models. You can create a folder, put all your custom upscale models in there,
and use them as you like. But in my experience, I never really needed anything else other
than what they have already provided. Now, here comes the question: if it is really
better than Lightroom or Chainer? In my opinion, it's better than both Lightroom and Chainer
combined. Upscayl has this simple workflow where you just upload your image, upscale
it, and walk away. You do not need to take any extra steps if all you want to do is upscale
the image. And if we talk about its capabilities, I already mentioned in my earlier videos:
ChaiNNer is the best upscaling model in my opinion out there. You can customize it any
way you want with a plethora of different upscale models, and all those upscaler
models that work in ChaiNNer also work in Upscale without ever needing to fiddle
with custom nodes for you to get there. But even if I say that this is the best AI
image upscaler, you should not take my word for it. If you want a detailed comparison
among all the popular image upscalers out there and decide for yourself, you
should watch this video right here.
