#Downsampled AA rendering
1 messages · Page 1 of 1 (latest)
oh yeah, that's much better
the thickness of the M differs a lot
hm, will need to look at each
Looks like the AA doesn't do what I hoped it would. I'd have to tinker
Anyway the performance takes a hit too, 2X pushes 4X pixels
0.5x: all three vertical strokes of the M in Mult seem like the same width with even spacing. however, the M in Common looks way off, and the N in Common is way too thin
1x: fixes a lot. interestingly, the two Ms in Common look different - left one seems uneven in the spacing of the "legs", right one looks even
2x and 10x look worse, actually, because the existing AA feels better than the "manual" AA
I have something that illustrates it better - just recording some footage to keep here for the future
I agree though
maybe take a look at font hinting too? that might???? help
This is 1X, you can see the font changes width awkwardly as the canvas moves slightly side so side
agreed
And this is 2X
there's no shimmering?????? that doesn't feel right
Well now the normal AA filter has subpixels to choose from
I can definitely understand it for the UI elements, but not the font
Like 2X looks worse to you in that video?
Honestly I like it when animations are a bit scuffed, but I don’t think that’s a common take lmao
2x looks better in motion for me
Yeah for sure the same with me
Objectively 2x looks better to me (but I like 1x a lot personally :)
But I still think I prefer 1X overall, I think I could figure out the other issues with 2X with some tinkering
I think LÖVE has font hinting support, so try messing around with that at 1x?
my assumption (with no experience with this lmao) is that font rendering already has AA, so maybe it's hinting which is doing weird things
Oh maybe yeah - sorry I was assuming I set it to nearest neighbour
for fonts
if I upres like this it'd have to be nearest neighbour for everything and then AA for the canvas
maybe that's also why