fallow iron
I'm making a shader which takes the RG channels of a texture and uses them as UV to look up colors from an another texture. For some reason the end result ends up having these white gaps in it, suggesting the shader looking up wrong coordinates despite them working perfectly in a Blender preview.
here's textures used
wispy epoch
I'm making a shader which takes the RG channels of a texture and uses them as UV...
at least make sure compression is disabled for the UV texture
fallow iron
here is the Blender shader result
fallow iron
at least make sure compression is disabled for the UV texture
all textures involved have compression off and no interploration
also tried converting color spaces between RGB and linear, didn't do anything helpful either
wispy epoch
That would have been my second guess. Does your shader in blender use linear color space too (and not sRGB)?
drowsy plover
Try to disable mip maps on the color texture.
fallow iron
they're not enabled
drowsy plover
🤔
sometimes the color sampling still does some gamma correction, you could try to put a color space conversion node between the sample output and uv input, and try linear -> sRGB or sRGB -> linear setting.
fallow iron
well that generally results in nonsense, but I'll try an offset
drowsy plover
well that generally results in nonsense, but I'll try an offset
linear > rgb is not better I guess ?
fallow iron
drowsy plover
I was also going to suggest a half pixel offset after this 😅
fallow iron
the offset was honestly my firstest solution long ago (this is an old bug, I just left it for later)
drowsy plover
Hum, the white bleeding seems to come from where different body parts overlap.
IDK how you are making the UV texture, but if you are rendering it from blender, maybe it's anti-aliasing the 3D objects (kind of expected) and blends the arm and body UV for example.
Disabling any form of AA in the rendering might fix it.
fallow iron
it's hand drawn pixel art
the UV maps are made by hand-placing pixels that match UV coordinates in the lookup texture
there's no AA involved as far as I'm aware unless Unity is hiding it from me
drowsy plover
What does the shader nodes look like in blender ? nothing particular I guess ?
I'm a bit cluelees about why it would behave differently now :/
fallow iron
I gotta find the most recent file because I lost it 💀
drowsy plover
I'm almost there though
fallow iron
👀
drowsy plover
The UV texture importer needs to have "Non-Power of 2" disabled
It was messign up the values.
fallow iron
how do I do that
drowsy plover
fallow iron
I don't see that being a thing on 6.2
is that a new setting or am I missing a package?
drowsy plover
I'm on 6.3, but it's an old setting.
Can you screenshot the texture importer ?
I've also added a half pixel negative offset to the uvs.
fallow iron
as far as I remember, the only thing the blender shader did past the basic mapping was the offset, can't find the most recent file right now
oh
drowsy plover
Oh, the texture is imported as 2D and UI, this is why you don't have the Power of Two setting.
fallow iron
it's because it was in Sprite mode
drowsy plover
If found a "fix"
I applied a "dilatation" filter to the color texture to avoid the white pixels.
Maybe in blender the texture was on a transparent background and blender was filling the pixels itself 🤷
And here's the graph setup :
viscid pewter
Afaik blender indeed does color dilation with alpha transparency automatically
Or the sampling method is otherwise somehow different enough to cause one pixel fringing to appear differently
drowsy plover
I didn't get the exact same result as the reference blender screenshot, so I'm not sure this is 100% correct though, but it's the closest I got.
viscid pewter
Mipmap sampling is also different, and something you don't really have control over