For 2D games it’s producing good “Fake\Pseudo HD” result. As far as I can tell it’s less wavy on slopes and has less false positives than others. Is it still true in 2024 or there are better more advanced alternatives?
SuperXBR is nice. It seems to try to preserve the original intent a little better.
Upscaling Pixel Art is another form of art.
There’s no consensus about how a low res pixel art should look in a hires display. So, there are many alternatives.
I, for example, think the results of AI upscaling applied for low res games are bad (for now).
Yeah, I realize that different people prefer different look. I’m specifically asking about “HD” or smoothing shaders.
Ask anyone to describe how Guybrush Threepwood looked in Monkey 2 from the main sprite.
You’ll gather so much different opinions that you’ll realize that the best way to correctly upscale is by using your own imagination.
My unbiased advice is to live everything as is, the way it was meant to be.
Add a bit of blur and let your brain do the hard work.
As a side bonus versus the trained and biased AIs is that it won’t eat megawatts for breakfast and still adapts to everyone’s needs for free!
from my experience with stable diffusion, most of the sprite ai uprezzing looks the exact same as xbrz/scalefx. Xbrz/Scalefx visually are exactly the same. Scalefx was made to have an internal scaler equivalent which didnt exist at the time. The xbrz external shaders are a little off compared to xbrz internal scaler. However, scalefx hit the xbrz internal scaler mark identically. Porting xbrz to an external shader after scalefx was made was a bit redundant. xbrz/scalefx is the successor to xbr. Xbr is now only relevant with the no blend variants when you want to rip sprites while keeping the low palette arrangement of the sprites. Xbrz/Scalefx add more colors to the sprites so a palette will keep 256 colors+. Shader forms could lag if a game is struggling w fps, causing the effect to come in late. Internal scaler forms stick to the sprites no matter what.