archived-shaders

[white pivot]

im applying effect in unity

[lunar depot]

A quick way to figure that out?

[neat hamlet]

I think checking if "_StencilComp" == 3 would work

[frozen fulcrum]

Hey! I have 2 game objects sharing same shader, is it possible to make surface inputs independent? Right now when i change for example float "Tint" same value automatically is changed in another object. Shader shares same material. Do I have to make 2 materials or there is another way?

[amber saffron]

You need two materials

[lunar depot]

But that would be on the material, not the component. Oh material != materialToRender while maskable is on might be an approximation

[rancid shuttle]

i really don't know anything about shaders but i tried to do double sided material and it always sets the color of mesh to white, even when i change it in material properties to any other color

Shader "DoubleSided/DoubleSided" {
    Properties {
        _Color ("Color", Color) = (1,1,1,1)
    }
    SubShader {
         Cull off   
         Pass {
             ColorMaterial AmbientAndDiffuse
         }
        
    }
}
``` can someone help?

[rotund tundra]

anyone know how to have a 2d int array (int[,]) in a compute shader?
i cant figure out how to define one

[strange roost]

https://www.youtube.com/watch?v=ERA7-I5nPAU In this video there are a bunch of shaders that emulate pixel art. I am not very shader savvy, so is there somebody who can deconstruct this?

[whole citrus]

https://stackoverflow.com/questions/16046293/do-directx-compute-shaders-support-2d-arrays-in-shared-memory

[rotund tundra]

the problem with this is that the size of the array in the shader is compiled, is there a way to set it dynamically as my array size is relative to the texture resolution

[rancid shuttle]

???

[frigid ether]

Thats pretty simple in reality : he uses a "cartoon" shader to have this lighting, and then he is rendering the whole screen in a low res texture, than he displays this texture using UI image on the screen

That's called a render texture : you put this in your camera output and you reduce the resolution

[trail nymph]

Hi, I'm very confused on what's going on

How do I properly sample _mainTex using the thread id?

I'm getting a Shader error in 'Test': cannot map expression to cs_5_0 instruction set at kernel CSMain at Test.compute(14) (on d3d11) at the moment

[strange roost]

What about the trees though? Towards the end of the video you can see in unity that each object has an individual shader, so how can it be one thing applied to the camera?

[frigid ether]

I'm pretty sure he does not uses an other shader for trees, seems like he just applied a very strong noise to the leaves to have this effect

You can see when he is in scene mode he does not have the pixelated effect

[strange roost]

Uh huh. So each material is just toon lighting with different colors, and the pixilation is added to the camera?

[frigid ether]

And he may have some noise distortion on his shader for leaves

[frigid ether]

I guess yes

You can see he has a material for each thing in his scene

[molten jasper]

It's a tad more complicated than reducing the render resolution and upscaling it, note the flat colours and lightened outlining in the mesh geometry

The tree leaf meshes could be the basis for a point cloud. The actual rendering isn't well explored in that video

[strange basalt]

the outlines look like they are baked on the rocks and stairs

looks like curvature data

[frigid ether]

Yep thats true

Would love to see his shaders for this

[strange basalt]

more or less a convexity map in addative

[molten jasper]

https://youtu.be/Lm5tRVNQxFs

[strange roost]

I'm asking because I'd like to translate this kinda style from a sort of orthographic perspective to full 3D. Would that work?

[molten jasper]

Explanation and link to the shader in the description

[frigid ether]

That might work yes, would be interesting to see

[molten jasper]

Reading through it, seems like it should work fine from a first person perspective

[strange roost]

I'd be having a 3rd person perspective. The project is along the lines of p5 or FE3H. Half social, half combat. The camera would be panned a lot closer, and the scale would be larger. It wouldn't really be a diorama as seen here

[molten jasper]

Either way it's perspective rather than orthographic, post what you come up with 👍

[strange roost]

Alright will do

that is if I'm able to come with something...

[frigid ether]

Ofc you'll be able ;)

[strange roost]

Oh uh one more thing. For more detailed things like houses, or maybe a sign with text, I'll obviously have to texture that manually. But won't that stick out like a sore thumb?

[trail nymph]

Edited the code a little, but still, all I'm getting is a black square...

[frigid ether]

Mh, maybe you could do a custom shader using the texture but still having the toon shading ?

[strange roost]

The texture will be pixel art still, so would the toon shader do anything?

[lucid scarab]

hi, I would need some help to understand a shader

I'm trying to convert it for URP with shader graph

[frigid ether]

Toon shader is more about how the light and shadows are working, so i guess yes

Never did toon shading on unity so I cant tell you much more about that

[frigid ether]

Yes, what' s the problem ?

[strange roost]

Ah alright. Well thanks a lot for the help. I'll see what I can come up with

[lucid scarab]

I don't understand this line
float2 uv_Texture = i.uv_texcoord * _Texture_ST.xy + _Texture_ST.zw;

[molten jasper]

I'd rather the font separately from everything else but still do depth checks

[amber saffron]

It applies the texture tiling and offset to the UVs

[lucid scarab]

_Texture_ST is populated by Unity maybe? it's only declared

[amber saffron]

Yes

[lucid scarab]

ok, cool. but this job is basically what a Sampler2D does...right? in shader graph

[amber saffron]

No, a sampler2D just samples a texture

Now, depending on the version of shadergraph, the texture object can contain the tiling and offset value, and pass it to the sampler automatically

[lucid scarab]

is this the case?

[amber saffron]

Does seem so

Are you able to use the "split texture transform" node ?

Or do you have a "use tilings and offset" toggle on the texture properties ?

[lucid scarab]

"use tilings and offset" is unchecked

[lucid scarab]

I can use that node

never used before

[amber saffron]

Ok, so if you check "use tiling and offset", you don't have to manually calculate it in shadergraph, and can just connect the texture property to the sampler to use it

[lunar depot]

Yeah, it does. I have a fix. I can access the material through materialForRendering but that has a bit more overhead than accessing the local materail so now I do this:

Not sure if GetFloat is better than .materialForRendering, tbh though

[lucid scarab]

what if I want to use the node? maybe tiling is xy from the starting example line?

[amber saffron]

Then use the "split texture transform" node, apply tiling and offset to the UV with the "tiling and offset" node, and connect the texture from split node + manually tiled UVs to the sample node

[lucid scarab]

I'm missing something 🙂

[undone anvil]

is its possible to make Depth of field auto focus from depth texture ?

well in case i can't do auto focus with RayCast hit to mesh collider instead read Grayscale value from depth texture

relate unity Forums // Its same Idea btw

[subtle mica]

How can i apply a top texture and a bottom texture with a shader graph

Is there a node to merge 2 textures

[regal stag]

Lerp, or Blend

[subtle mica]

How can i apply this?

[regal stag]

In this case you'd probably sample both textures with separate Sample Texture 2D nodes.
Then blend/lerp between them based on the Y axis of the Normal Vector.

[subtle mica]

What's wrong?

there is no hard cut, the textures get mixed

Nvm i got it

You have to use Lerp

[amber saffron]

I don't think the preview displays the timing and offset, you need to apply this to a material

[amber saffron]

It is technically possible.
The best way is probably to come use a compute shader to get the focused depth, and pass it to the DoF effect

[lament scarab]

How could I modify this shader so that there is no horizon (grey part). I know for a fact that it's something in the shader code that does it, since it's not the default behaviour but I can't understand shader code well enough to fix it.
https://github.com/n-yoda/unity-skybox-shaders/blob/master/Assets/SkyboxShaders/Skybox-StarrySky.shader

[tall girder]

how can i make a shader not overwrite the default texture of my mesh?

[grizzled bolt]

In what context?

[tall girder]

on the left is the default texture, the right is my shader. I want to get the texture from the mesh and apply a simple animation to it

[tall girder]

you see, sampletexture 2d and manually setting the texture isn't an option here....

[grizzled bolt]

Not sure what I'm looking at here, why's there so many of the same texture?
Anyway, when you switch material for the effect I assume you would be able to grab the texture reference property and pass it onto the effect material at runtime

[tall girder]

i found a way i think. I can set 2 materials in the mesh renderer.

[twin raptor]

Will compute shaders written with legacy RP work in URP/HDRP?

[tight phoenix]

does anyone know why the result of this graph tiles across 4 of the faces continuously, but is split separate on the front/back on one side? Is it a UV thing?

is it supposed to do that?

I would expect it to not tile across all 6 faces, or tile across all 6 faces, but not tile across some faces but not others

because its doing that, my end result looks bad

[regal stag]

Yeah I think this would be the UVs (well, the mesh tangent vectors are aligned to the X axis of the UV coordinates). The cube is probably unwrapped using a "lying down cross" shaped net, which might explain why 4 of the faces look similar

[tight phoenix]

as far as I can tell two of the six faces are just backwards for no reason and I cant move forward until I can fix it

[tight phoenix]

if that is the case, where can I replace in the graph to counteract this? At no point in the graph is the UV sampled which confuses me

the out is plugged into the albedo directly atm

Ive tried removing and switching different pieces but it never removes that 2/6 faces weird brokenness

[regal stag]

I'm not too sure. You could just use a Vector3 node rather than the Tangent Vector, but then you won't get the gradient wrapping around the edges of the cube...

Might just need to unwrap the cube mesh differently. Maybe having all faces in a line next to eachother.

[tight phoenix]

to the best of my knowledge the default cube maps its UV on all 6 faces copy paste over eachother

like its not a long line

[tight phoenix]

I will try making a new mesh like that real fast to see if that fixes it, but if it does, I dont know how I can use that fix because that will mess up a lot of other things

I dont understand how the UV is affecting it

because the UV is never sampled

shouldnt UV only affect it if it samples the UV?

none of the nodes pass in its UV channel 0, 1, 2, any

if it never samples the UV, how does it know?

[regal stag]

As I mentioned above, the Tangent Vectors stored in the mesh are aligned to the X direction of the UV coordinates.
(And it's aligned like that, because that's basically the purpose of the tangents - to allow converting between World and UV space, mostly for normal mapping)

[tight phoenix]

hm so you're saying like, two of the 6 faces are rotated wrong?

[regal stag]

Am wondering if using the Normal Vectors instead would produce better results

[tight phoenix]

ill try exporting after rotating two of the faces, and then try that

[regal stag]

Yeah, I would assume two of the faces are turned 90 deg in the UVs

[tight phoenix]

hmm, rotating two of them didnt change a single thing

ill try rotating another two in case I just coincidentally didnt do it right??

okay the first time must have been coincidence that it kept working

the second time affected it

hrmgh im not sure whats the better sollution now though, because its a cube, wont it be impossible to make it tile across all faces?

right now it tiles across no faces 🤔

and im not sure if thats better

[tight phoenix]

no matter what I do it just keeps getting worse

[tame topaz]

@tight phoenix We don't need the monologue

[slate compass]

what in the flying fuck

someone just lost their shit in front of the internet

[hushed wing]

Last few days as well

[tight phoenix]

im not really surprised

i have nothing so its not like I can go any lower

this is the usual reponse to someone who is in crisis, tell them to fuck off because no one wants to see someone struggling

[slate compass]

lmao.. our job is to help with game engine problems

not to console ur craziness

[hushed wing]

I mean, this is a dev server. Not a crisis server.

[tight phoenix]

everyone i reach out to does this, so i never get any support, its always people brushing me off because its not their fault or responsbility or whatever

yeah i know that

[slate compass]

then go get help in the appropriate places

[tight phoenix]

amyways talking to you isnt helping so im closing discord

[slate compass]

see ya man, get help

[wary horizon]

perhaps you don't get support because the last 75 messages, cannot be helped with

[tight phoenix]

k

thanks for letting me know i should block you 👍

[wary horizon]

😢

[tame topaz]

@tight phoenix This isn't a mental health or crisis support community and we're not equipped to be one. If you need help, please seek it, but otherwise please keep this channel on topic to shaders and shaders only.

[karmic hatch]

I think the tangent vector and UVs may be related to some extent (one as the derivative of position wrt the other perhaps) but the main thing is you either want to switch from diffracting lines to a grid of diffracting elements (just copy again, but this time with normal cross tangent vector as your input), or you could switch from tangent vector to 1/2 (reflect(vec3(1,1,1), normal) + vec3(1,1,1)) which should be continuous (though diagonal across cube faces)

anyway, it's very often that code does something wacky you don't expect, and often shaders are a pain to debug because you can't get out numbers easily. It happens to everyone.

[wicked niche]

I use this method, but i dont have a rectangle

[molten jasper]

I need to change a fragment colour based on the central view direction and the normal of the fragment. How would I go about doing this?

I have a gradient, I'd like for it that any normal facing left to be t=0 and any normal facing right be t=1. where both left and right here mean perpendicular to the view direction and surface normal

[shadow locust]

project the normal onto the plane perpendicular to the view

then just check the x direciton of the projected vector

[karmic hatch]

Get the normal in view space, and then return the x component (will need a remap too)

[molten jasper]

This is what I have thus far, I hope this screenshot is legible

I'm aware that sharing screenshots isn't usually helpful as it requires the recipient to parse what's going on. Shaders are not my forté and I'm fairly lost when it comes to graphics programming

I'm hoping to achieve a sort of pearlescent/iridescent effect using the pastel blue and pink

[karmic hatch]

I think you can cut out a lot of it by using a Normals From Height node

Plug in the gradient noise and you can remove everything up to and including the Normal Strength node

[molten jasper]

it doesn't quite give the same result

[karmic hatch]

Looks like it's just a difference of normal strength

you should be able to pick it so they're the same

[karmic hatch]

One is stronger than the other but if you change the value going into normal strength you should be able to get them identical

[molten jasper]

I did change the strength here to be roughly the same

[karmic hatch]

it's probably the same under the hood, just one saves space on the graph and is clearer imo

[molten jasper]

yeah

[karmic hatch]

The implementations might be slightly different so the same value input for strength yields different results, but it's probably just a factor of 2 or sqrt2 or whatever off

so if you just take the value in one of them and change it you could get it to look identical to the other

[molten jasper]

sure, this is good to know

[karmic hatch]

Anyway once you get the normal, put it through a transform node to take it from tangent space to view space. Then remap the r value from (-1, 1) to (0, 1) and that's your gradient input

[molten jasper]

bizarre

[wicked niche]

drew such a rectangle, but now how to make smooth for the edges of the rectangle through smoothstep, please tell me

like this

[karmic hatch]

Rather than using an if and bools, use a distance function (Inigo Quilez is the go-to for basically any distance function; he's got them all). Then plug in the distance function to your smoothstep (i would also put a remap in between so you can change how smooth/sharp and rounded it is)

@wicked niche

[rotund tundra]

i have a colour as a float4
how do i set a specific pixel using x,y on a RWTexture2D to that colour?

[dim yoke]

TheTexture[int2(x, y)] = whatever i think

[west locust]

i am developing a game for mobile and i am using a "Mobile/Particle/AlphaBlend for trees and fence but it show object trough each other?

and the textures is a PNG

[cosmic prairie]

use alpha cutout for fence

are the trees baked into a single mesh?

[west locust]

i have created the level in blender and trees are all one object

but the texture is same for both fence and trees

[amber saffron]

Are the fences a single big object ?
If yes, what you are seeing is transparent rendering order making that the fences are drawn after the trees, when the pivot of the object is closer to the camera.
Either split the fences into smaller sections, or like mentioned before, switch to alpha clip opaque

Same can happen if the trees are a single big object of course

[west locust]

so if i separate trees into single object and divide fence into smaller piece will this solve the issue?

[amber saffron]

Likely

[wicked niche]

ok, what function can be used for a rectangle, so that later it can be correctly used with the smoothstep function

[thin crater]

Hi, is there a way in the shadow graph URP to receive shadows for an unlit shader?

I'm trying to create a grass shader, and if I use a lit shader type only one side of the plane that represent grass reflects the sun, the other side don't, so the unlit shader has the best effect since it has uniform colors for all the grass

But it doesn't receive shadows..

[karmic hatch]

Inigo Quilez has a distance function online

[mental bone]

how would I do the texture projection described at 38:34 https://www.gdcvault.com/play/1012264/Shading-a-Bigger-Better-Sequel in shadergraph ?

[thin crater]

Another solution could be using a lit shader and receiving light from both sides. Can I do that? Is there a way?

I don't know why a plane with rendering face set as both would reflect light only from one side...

Depending on the rotation, as you can see from the different color on the left in the same mesh, it reacts to light differently...

[regal stag]

The matrix (or perhaps just an offset) defining the ellipsoid would be sent into the shader. You'd use the fragment position (Position node (object) space) and transform that using that matrix (Multiply). Or in the case of an offset just Add/Subtract.
Then Swizzle (e.g. "xy", but depends on which axis you want to project from) and sample using that.
It's not too far off the "worldspace planar mapping" idea, (e.g. https://www.cyanilux.com/faq/#sg-planar-mapping) but in this case it's just object space (or "ellipsoid space"). You'd also want to make sure the texture is set to Clamp rather than repeat, and the edges of the texture should be black/transparent to avoid clamping colours at those edges.

[regal stag]

Shadows in unlit graph : https://www.cyanilux.com/faq/#sg-unlit-shadows
Shading back faces in Lit graph: https://www.cyanilux.com/faq/#sg-two-sided-shading

[thin crater]

wow ok

thank you

which one do you suggest?

[regal stag]

Another common method for grass is to override the normal vectors to only point upwards, (0, 1, 0) in world space

[thin crater]

How could I do that?

[regal stag]

As in, connect a Vector3 node with those values to the Normal port in the vertex stage.
Or the one in the fragment stage, if you change the "Fragment Normal Space" in graph settings to use World space.

[thin crater]

Like that?

It didn't work

As you can see from the left plane that's light green and the right that's dark green, one of the sides is still getting light

[regal stag]

Oh right, the vertex port is in Object space, I forgot. You'd also need to use a Transform node from World to Object (Direction mode) for that port

[thin crater]

Well, you sir have resolved in a moment what many guys that tried to help me could not

Thank you so much

[regal stag]

If you combine the planes in a single mesh you could also continue using (0,1,0) - or whatever axis is up - in object space so that rotating still affects the lighting of the whole object

But since grass tends to point upwards in world space, this works too

[mental bone]

I recreated this in SG. Seems to be what you are talking about but the texcoords simply color my entire model they dont seem to project the texture at all

[thin crater]

It should never point in any other direction so that should be ok

I've got another question, probably a bit harder

another of the ones where noone could help 😛

I looked for a custom biome shader solution for my terrain (subdivided in chunks) for a lot of time. In the end I found nothing and created my custom shader: it takes a texture mask as an input, a texture that will be applied where the green part of the mask is and it lerps this with all the biomes.

So that's the single biome node:

And repeated and lerped 4 times, I get my shader:

It works kinda well, but I've got a bug and something I'd like to change, if you can help me... The bug is, borders of the biomes, if seen from far away, are black

[regal stag]

Personally I would convert all these Vector3 inputs and Dot Products into a Matrix4 and Multiply, as it'll make the graph much simpler. If the ellipsoid is controlled via a GameObject/Transform can just use the transform.localToWorldMatrix for it.

You also aren't sampling a texture with these coords yet

[thin crater]

That's what is happening

if you look where the biome changes (from grass to forest grass) you'll notice a black line...

Another example in a beach, where it's easier to be seen:

[regal stag]

Unrelated to the problem, but assuming that Color Mask node is using green (0,1,0) you could remove that node and just use the G output from the texture sample. Would be cheaper.

[thin crater]

(there are two lines because the sand is actually two biomes, seafloor is the one close to the water and the other is normal sand, close to the land)

[thin crater]

Yeah, right. Removed that, gonna test it soon

[regal stag]

I feel like for the first texture you don't need to do the biome masking. You're lerping from (0,0,0,0) which is probably where the black is coming from.

[thin crater]

ok it worked perfectly well

[mental bone]

I am sampling a texture with the interpolated coords down the shader. The texture is set to clamp as well. I'm just plugin them into the UV of the texture sampler. Should I be doing something mroe with them ?

[thin crater]

What could I use instead of lerp? Multiply?

I'm a real noob to shaders (your website seems like a paradise to begin experimenting with it)

[mental bone]

I copied the code from the presentation

[regal stag]

The lerps are fine, I just think for the first one you could skip that and put the texture sample result straight into the A input on the next lerp. Assuming you only have 4 biomes?

[thin crater]

Yes

Every time I need another I just copy and paste it

the mask is then rendered via c# code, and that's what I wanted

[regal stag]

Then yeah, try skipping this first one. If all the other biomes are masked on top you don't need to mask this one too.

[thin crater]

I'm trying to test what you suggested, editor is stuck into importing 😛

8 gb ram is not enough for unity

[regal stag]

You could also look into combining the masks into one texture and use the other outputs rather than just the G channel. Would use less texture samples that way. It's similar to how terrains handle splatmaps.

[thin crater]

It worked perfectly. Thank you so much

No more black lines

[regal stag]

Okay, didn't see that in the graph you shared above but afaik it should work then. Maybe check your Vector3 inputs are being set properly

[mental bone]

I'm manually setting them as exposed properties

the cliping part seems to be working, this is the whole graph at the moment

[thin crater]

Combining it via shader? With a ..combine node?

[regal stag]

No, I mean in C# or however you create the masks (hand drawn?)

[thin crater]

C#

(It's a procedural world)

So I create a big texture, or a texture array, and feed it to the shader?

Anyway optimizing it is probably something I'm gonna do in the future, right now it works pretty well and I don't even know why ahah

If I can, I'd like to ask the final and probably hardest question 😛 one that has bothered me for about two weeks before just giving up

[regal stag]

The texture can stay the same size, just use the other colour channels. Red, green, blue, alpha, for different masks

[thin crater]

Ah, right, I tried it

but I'm gonna need many more biomes, currently there's four but probably they're gonna be around 10/15 with Version 1

Thats also why I didn't use vertex colors, iirc (I coded this shader three months ago)

I might filter colors...

not sure about that

[thin crater]

My biggest failure with my project are the biome edges, I spent so much time trying to get a system that "smooths" those edges, but I never found any way to do it

[trail nymph]

Hey, anyone here know anything about compute shaders?

[thin crater]

My main goal was to use some sort of lerp between them to get a value of how much close the point was to the edge of the biome, and then use a noise map to do some sort of biome-crossing effect

[regal stag]

Difficult to debug this. What is the splat texture like? Is it greyscale or transparent? If it's transparent you might need to use the A output from the texture sample instead

[thin crater]

A bit like that, but at a very small scale (smaller than a "block" that you can see in the image, so each block would be clearly one biome or another)

[regal stag]

I guess the mask textures are using Point filtering? Could use Linear to get a bit of a transition.
Maybe also upsampled/downsampled to adjust that, or use a smoothstep in the shader after sampling the mask

[thin crater]

I gave up after a week since I couldn't even get close to something like that 😦 then I tried to have a "smooth" biome change, where edges were a bit less hard and more rounded.... Gave up on this too

[thin crater]

Yes

I tried that - but I don't remember why it didn't work. I'm gonna do that again and see why I didn't use this solution

[mental bone]

It is indeed transperat, using the alpha turns the entire model black, so the problem is with the uv's I think

[regal stag]

It might be the alpha clipping part is hiding the texture now?

Maybe adjust the value in the Alpha Clip Threshold port

[mental bone]

hold on I'm seeing something, It might turn out that my elipsoid was badly defined

I need a way to visualize it heh

[thin crater]

Well it simply didn't work, I've set it both to bilinear and trilinear and nothing changed

[regal stag]

Oh, you have a SamplerState node in the shader, did you change that? Or the settings on the texture?

[trail nymph]

I'm processing a big image in my compute shader and I want to count how many white pixels it has. I've tried making a RWStructuredBuffer<int> and in the CS incrementing the first index if pixel == 1, but after reading the data the values are in the 50-64 range, seems like the thread groups overwrite eachother. What'd be the best way to do this?

[thin crater]

Settings on the texture, via script

second line

[regal stag]

Okay then yeah, you'd also want to remove those SamplerState nodes, as it'll override it

[thin crater]

Ok this simply works and I'm astonished because I tried the same thing that you made me do over and over without success

BUT there's a bug 😛

It always belnds every biome with the first biome on the shader, the one where I removed the lerp

[regal stag]

This thread might help, answer uses InterlockedAdd
https://forum.unity.com/threads/solved-how-can-i-count-the-number-of-pixels-above-a-threshold-with-a-shader.763232/

[thin crater]

So for example between those two biomes (sand and seafloor, I use it to not spawn beach objects too close to the sea) the blending has grass in between

[trail nymph]

Oh hell! Big thanks let me try it out

[thin crater]

Also, I recalled the reason why I didn't use this solution: sometimes behaviour between chunks is weird

like there, you can see an hard edge in the center, where two chunks touches

[regal stag]

Mmmmmm yeah. This is where combining the masks into each colour channel would really help 😅

[trail nymph]

Worked like an absolute charm, tysm

[thin crater]

if I can fix this I'll probably maintain this solution, since it's better than hard edged biomes

[regal stag]

Ah yeah, that's also going to happen unless the mask is also padded with values from neighbouring chunks. Would then need to scale the UVs a bit so you aren't sampling those parts

[thin crater]

but if you have any suggestion... Or if you know how to make a "blending" effect with the noise, as mentioned above 😛

[thin crater]

So I'd have to give my mask a +2 on size (one left and one right), and then scale it until those two are hidden?

[regal stag]

Yep

[thin crater]

That's pretty hard because biomes can be modified from the player (and saved in a file), so I'd have to load neighborhood chunks data to get what biome is the closest one.. And chunk data loading takes a lot of time 😦

[regal stag]

Blending with noise is also doable. Can distort the uvs used to sample the mask.
Would probably use Gradient Noise node (or a seamless noise texture) with worldspace planar projection (https://www.cyanilux.com/faq/#sg-planar-mapping) - this'll make sure the noise doesn't cause seams over each chunk
Subtract 0.5 to "center it", Multiply by a small value (e.g. 0.02), then put this into the Offset port on a Tiling And Offset node, which would go into the UV port on the mask sample.

[molten jasper]

Can I do fresnel based on the normal map?

[regal stag]

Sure. Sample the normal map, then use a Transform node to transform it from Tangent to World space (Direction mode), for the Normal input on the Fresnel Effect node.

[molten jasper]

iiiinteresting. I think this will hopefully give me iridescent effect I'm hoping for

[mental bone]

its sorta working, but turns out I have to idea how to define a elipsoid

[thin crater]

Well.. Without the worldpace planar projection, this has an interesting effect, but that's not quite what I was looking for...

[thin crater]

I hoped to have an effect more like this, or at least with borders not so squared (round borders with this would be ok, I think)

So, using gimp and a lot of shame for the quality of what I did, the output would be something like that...

no hard edges

is it actually possible?

[regal stag]

I think the noise scale here is just too small

[regal stag]

But adjusting the scale is unlikely to give this result. It'll still look blocky, just... a bit wiggly too

To make it look like that you'd probably need a higher resolution mask texture.

Or at least something with more of a gradient in the masking. With linear filtering and a step, that can produce nice results. Same idea used in SDF textures. (e.g. https://twitter.com/cyanilux/status/1188793223029886976)
Might be hard to get that gradient though, without applying blur operations or jump flood algorithm to generate the sdf.

[thin crater]

I tried to increase the mask texture, computation time went really high. Anyway I need to have a "block" about this size, for gameplay purposes

[thin crater]

Gonna try that.

[rotund tundra]

so im learning compute shaders, and as i understand the goal is to divide the task between the threads of the shader, so with 8,8,1 numthreads you divide the task 64 ways.
is that understanding correct?
if it is im wondering a few questions
are the threads 0-7 or 1-8 and
how do i devide a for loop between the threads correctly?

[thin crater]

Have you got a tutorial about linear filtering?

Using steps and trilineal filtering has an effect really close to what I'd like to have 😄

[mental bone]

@regal stag I tried the matrix approach you suggest, I fill it from a helper object like this renderer.material.SetMatrix("_ElipsoidMatrix", transform.localToWorldMatrix); but it does not work as expected, it only works if I scale the objec to something very big like 40

[molten jasper]

Ok so fresnel will give me a gradient effect but I need something directional relative to world directions

ignore me, just came up with an idea

[thin crater]

Using steps also fixed the in-between biomes grass.

I'm going to see how it goes between chunks, and with single square biomes (e.g. when the player modifies the terrain manually, you could have a single squared biome)

[regal stag]

Maybe that's because the model it's being applied to has some weird scaling?

[thin crater]

if that works, I'm probably gonna cry lol

[mental bone]

nope its scaled 1,1,1

[regal stag]

Hmm maybe localToWorld isn't what you need. I guess it needs to be relative to the parent object...

Maybe enemy.worldToLocalMatrix * ellipsoid.localToWorldMatrix ?

[mental bone]

renderer.material.SetMatrix("_ElipsoidMatrix", renderer.transform.parent.transform.worldToLocalMatrix * transform.localToWorldMatrix); it seems to be inverted ?

[karmic hatch]

afaik yes that is correct.
I believe threads are 0-7
Give each element in the loop its own thread, then store the result back in the buffer. It doesn't work if the loop is sequential though, it has to be parallelisable.

[rotund tundra]

its a 2 dimensional for loop over a RWTexture2D

[karmic hatch]

or a buffer

[olive aspen]

Hello guys, I have run into a problem, so I followed carlikecoding's tutorial on compute shaders and I'm pretty sure I did everything properly but when I entered play mode it said that I was giving very expensive calls to the gpu

Here is the compute shader:

RWStructuredBuffer<float3> _Positions;

uint _Resolution;

float _Step, _Time;

float2 GetUV (uint3 id)
{
    return (id.xy + 0.5) * _Step - 1.0;
}

void SetPosition(uint3 id, float3 position){
    if(id.x < _Resolution && id.y < _Resolution)
        _Positions[id.x + id.y * _Resolution] = position;
}

#define PI 3.14159265358979323846
float3 Wave(float u, float v, float t)
{
    float3 p;
    p.x = u;
    p.y = sin(PI * (u + v + t));
    p.z = v;
    return p;
}

[numthreads(8,8,1)]
void FunctionKernel(uint3 id: SV_DispatchThreadID)
{
    float2 uv = GetUV(id);
    SetPosition(id, Wave(uv.x, uv.y, _Time));
}

And here's the script responsible for the graph: https://pastebin.com/RC9JP5yZ

It should show a lot of cubes in the center

[thin crater]

It works so, so good! I'm really really happy. The only bug I found is in between chunks

As you can see it applies the grass texture even there

[olive aspen]

This is the error I get

[rotund tundra]

for(int x = floor(Size / id.x); x < floor(Size / (id.x + 1)); x++)
    {
        for(int y = floor(Size / id.y); y < floor(Size / (id.y + 1)); y++)
        {

does this look right?
size being the texture resolution

[thin crater]

Sometimes it simply spawns green grass (forest grass) where probably the mask doesn't reach one or another. Those dark green spots are from the grass texture, the one without the lerp

[karmic hatch]

is this in the shader?

[rotund tundra]

yes

it doesnt crash but it also doesnt work

[karmic hatch]

In the shader you get a single id(x,y,z), and that's your int x, int y, and int z (if you use one)

[rotund tundra]

yep

[karmic hatch]

Don't loop over everything bc then you're telling the shader to, for every element in the array, loop over every element in the array again

which is going to be very expensive indeed

unless I'm misunderstanding what that loop is for

[rotund tundra]

each thread should do 1 64th of the loop?

i think

im still really new to multithreading

[karmic hatch]

Basically, on the CPU you'd have a loop over x, and inside, a loop over y.
The GPU already gives each thread a value for x and y (id.x, id.y) and then you use those values to calculate whatever you needed to calculate, put it into its slot on some big array, then hand it back at the end

[rotund tundra]

// for(int i = 0; i < Size; i++){
    //     for(int j = 0; j < Size; j++){
    //         if(pow(i - PixelCoords.x, 2) + pow(j - PixelCoords.y, 2) < pow(SplashSize, 2)){
    //             //splatMapData[i, j] = team;
    //             if(Team == 1){
    //                 InkTexture[int2(i,j)] = AlphaColor;
    //             }
    //             if(Team == 2){
    //                 InkTexture[int2(i,j)] = BetaColor;
    //             }
    //         }
    //     }
    // }

this is what im trying to implement properly
it renders a circle of radius SplashSize, around PixelCoords

[karmic hatch]

So on the CPU, it'd be

Vector3 SomeFunction(int x, int y)
{
    // stuff
}

for (int x = 0; x < numX, x++)
{
    for (int y = 0; y < numY; y++}
    {
        bigArray[x, y] = SomeFunction(x,y);
    }
}

On the GPU, it gives you x and y - you can think of it as already doing the looping for you - so you just need

Vector3 SomeFunction(int x, int y)
{
    // stuff
}

bigArray[id.x, id.y] = SomeFunction(id.x, id.y);

[rotund tundra]

uhhh

wait

[karmic hatch]

obviously I'm bastardising the code a lot here but the point is the looping is done for you and you get id.x, id.y as your x and y equivalents, so you now just need to run whatever function with them as inputs

[rotund tundra]

[numthreads(8,8,1)]
void CSMain (uint3 id : SV_DispatchThreadID)
{   
    InkTexture[ int2(id.x, id.y)] = GetColor(id.x, id.y);
}
float4 GetColor(int2 PixelCoords){
    if(pow(x - PixelCoords.x, 2) + pow(y - PixelCoords.y, 2) < pow(SplashSize, 2)){
            if(Team == 1){ return AlphaColor; }
            if(Team == 2){ return BetaColor; }
        }
}

so like this?

[karmic hatch]

yes

[rotund tundra]

wait

but wont this run 64 times

my texture is 256

[karmic hatch]

Each thread group will contain an 8x8 block of threads, and when you set it off, you tell it how many thread groups it needs to generate and run

So it'll run the right number of times

[rotund tundra]

uhh

so the dispatch uint3?

[karmic hatch]

You tell it how many groups to create in the Shader.Dispatch() on the c# side

[karmic hatch]

yes

[rotund tundra]

ok

uhh

wait

its now rendering a quarter of the circle

does that mean my Dispatch size is wrong?

[karmic hatch]

I think so, what dispatch size did you give it and how many pixels is your texture?

[rotund tundra]

8 * 256
texture is 256

[rotund tundra]

wait

[karmic hatch]

Is the texture 256x256 or 16x16?

[rotund tundra]

256

so it should be size / 8

[karmic hatch]

yes for a 256x256 texture with 8x8 threads, you should need 256/8x256/8 thread groups, or 32x32

so uint3(32,32,1)

[rotund tundra]

still getting quarters

[karmic hatch]

oh pow() doesn't like negative numbers

abs the input

or better yet just multiply things by themselves instead of using pow bc iirc it's slightly faster that way

[rotund tundra]

now its not rendering anything

[karmic hatch]

what did you change it to

[rotund tundra]

wait my bracket were wrong

OK WE GOT CIRCLES

[karmic hatch]

ayy nice!

[rotund tundra]

alr how do i make it not overlap or clear the last render

[karmic hatch]

What's the shader displaying the texture?

[rotund tundra]

[numthreads(8,8,1)]
void CSMain (uint3 id : SV_DispatchThreadID)
{   
    InkTexture[int2(id.x, id.y)] += GetColor(PixelCoords, id, SplashSize);
}

[karmic hatch]

The texture goes onto a quad though right?

[rotund tundra]

yes

oh

its just unity standard

[karmic hatch]

Transparent, right?

[rotund tundra]

cutout

[karmic hatch]

Wait, there are multiple circles per quad?

[rotund tundra]

yea, the circles should add to the texture, but not stack

[karmic hatch]

Swap it from += GetColor to = GetColor

[rotund tundra]

that clears it when adding a new circle

[karmic hatch]

ah right

= Max(InkTexture[], GetColor())?

or GetColor.a == 0 ? InkTexture[] : GetColor

(assuming you're using the alpha channel for transparency)

[white pivot]

i want my effect to fill the image what node i must input before it?

[rotund tundra]

i made it pass the old colour through GetColour and return it if it didnt need to update

[karmic hatch]

Nice

[rotund tundra]

Thank you soo much

[karmic hatch]

np :)

[tight phoenix]

tried various methods to calculate tangent vector but none of them duplicated the unity tangent node
trying to copy what you said to do in this post now

this is all x/y, all I really want to do is fix my cube so that all the anisotropic lines go in the same direction

[tight phoenix]

normal cross tangent doesnt fix.

[tight phoenix]

your second suggestion 1/2 (reflect(vec3(1,1,1), normal) + vec3(1,1,1)) also does not work, they all go in different directions

[tight phoenix]

it would help if you can clarify these two formulas, ive implimented them as written to the extend of my understanding of them

[karmic hatch]

If you combine it (blend, add, whatever) with the original effect (just using the tangent vector), the combination should be face-independent again (or so I think)

[karmic hatch]

you're running up against the Hairy Ball theorem, which basically states that on a strictly convex shape, you can't have a vector parallel to the surface everywhere that's also continuous everywhere. You have to have a pole where something either points up/down, or swirls in a spiral (discontinuity). On a cube, you could hide it by putting the discontinuity/sticking up at a corner (which is what I was thinking of with the vec3(1,1,1) thing)

[tight phoenix]

hmm hairy ball theory, I see after googling it. So I was trying to fix something unfixable

[olive aspen]

Hello guys, I'm following this tutorial and when I get to the end of 2.1, if I do a high value in the resolution (like 200) Unity crashes for me

[tight phoenix]

Ill try your fixes here to run it into a corner

[olive aspen]

I don't know why it does that, I've literally copied every single line of code

Anyone?

[tight phoenix]

this is sorta working, except that some of the faces are super washed out bright while others are double dark 🤔 still trying to normalize that

oh right, normalize will literally do that

i guess this is as good as it can get, since hairy ball theorem is unsolvable

[tight phoenix]

do you think this is fixable? straight on from certain faces, its dark, while others are bright

how hard would it be to aim all the same corner instead of arbitrary corners?

[tight phoenix]

I do have set of UVWs that point from one corner to the other, could they be used in some way?

[tight phoenix]

certain faces are very bright and others are very dark

I cannot solve this on my own so Im going to drop it now

[molten jasper]

Looks cool

[tight phoenix]

Thanks, I am not satisfied with it, it looks 80-90% of the way there but its still not quite 100%

[molten jasper]

Keep going

[tight phoenix]

through experimenting realized that is the bitangent

[karmic hatch]

Sry I mean combine the result from putting in tangent vector, and putting in (tangent vector cross normal)

Not just the vectors

[tight phoenix]

Im not sure I understand, are you saying to duplicate the entire graph?

[karmic hatch]

Yes

[tight phoenix]

like the whole thing?

that doesnt seem very performant

[karmic hatch]

Indeed

I think a real diffraction grating would be multiply, add could correspond to two mostly transparent gratings or something

[karmic hatch]

Though it depends how expensive the original is

I think there's probably a way to align it diagonally and have it all work the first tims

[tight phoenix]

funnily enough the result of the above seems to be exactly the same as https://cdn.discordapp.com/attachments/497874081329184799/1014574164457308260/unknown.png using the combined thing at the start

well not exactly but visually hard to tell a difference

[tight phoenix]

I do have UVs https://cdn.discordapp.com/attachments/497874081329184799/1014565037748273272/cornerUVs.gif from one pole to the other

but I was not able to figure out where to input them to get the result desired

https://gfycat.com/repulsivegreatbuffalo

maybe this is good enough and I should stop sweating it

I think its the inclusion of the main light direction that is causing my grief

its trying to simulate brightness but that brightness is going to come from being a lit shader

removed the main light portion and let the actual main light light those surfaces

I am not happy with it and would prefer it to be perfect though

it looks fine but its not perfect, and i struggle to accept anything less than perfect, but I don't have the ability to obtain perfect

its traumatic for me to give up, I can't give up, I don't want to give up, my anguish comes from chasing the perfect appearance, and my inability to achieve that look

im already falling into the same crisis as yestrerday openly admitting this so I am going to shut off my pc now

[molten jasper]

Oh goody, anywhere to report this?

[grizzled bolt]

https://unity3d.com/unity/qa/bug-reporting

[jade edge]

so im having trouble with a mesh

and i cant find a thread for those things

where would i put it

[dim yoke]

Mesh? Procedurally generated? What trouble? (Doesnt sound shader related atleast, you can ask on #💻┃unity-talk if you dont know any better channel)

[jade edge]

@dim yoke its generated by code

so procedurally yes

[quick cape]

How can I make a Util class that I can use from .shader?

[regal stag]

Can make a .hlsl (or .cginc, more common in built-in rp) file. Include it in the shader via #include "Assets/PathToFile/Name.hlsl"

[quick cape]

Great, thank you!!

Would you recommend either .hlsl or .cginc over the other?

[regal stag]

Don't think it matters really, they are the same thing

[quick cape]

Oh my gosh, that was easy!! Thank you again!!!

[molten jasper]

Is this right? I'm sort of losing the lavender colour

[dim yoke]

Then either on one of the coding channels beginner/general/etc. or #💻┃unity-talk depending on the nature of the question

[jade edge]

its always nice when two ppl tell me to talk to the other one

[regal stag]

I'm not sure you should even have the lavender colour when using the Normal Blend node. If they're normal maps should sample using the Normal mode to convert it to an actual vector (like -1 to 1 range). Or use Normal Unpack.

[molten jasper]

The shader graph doesn't seem to like using -1

Hopefully it's roughly clear what I have going on here

I've separated the edge bevels out but need to do the normals for them. What's the proper wait to detail them?

[regal stag]

Hmm. I'd probably do something like right - left and top - bottom (referring to the edge bevels here), and put them into the X and Y inputs on a Vector3 node. Z of 1. Then Normalize.

[molten jasper]

It needs that sort of base to subtract from otherwise it ends up like this

[regal stag]

That might be okay? The previews don't show negative values so hard to tell. Should connect it to the Normal (Tangent) port on the master stack fragment and see what it looks like when shaded

[molten jasper]

It's a tad strange

top and bottom were the wrong way around

[thin crater]

Hi! I tried to add foam to my water. In the scene view, it's perfect, but in the play view I see nothing. Any clue why?

Scene:

Game:

The camera renders to a render texture and not to a display. Might it be it?

[regal stag]

Depth Texture is probably not enabled on URP Asset

[thin crater]

It was, I enabled it but nothing changed

Camera is also set to get this option from urp asset

[regal stag]

Might have multiple URP assets per quality setting?

[floral barn]

has anyone done a water shader in built in unity 3D before (not urp/hdrp)

[thin crater]

I have (all the different quality urp settings), but I double clicked the right one from the quality panel so I'm pretty sure that it's the correct one

[regal stag]

Idk, I'd try enabling it on all of them

If that doesn't work then maybe the Frame Debugger window can give some hints

[thin crater]

Enabled on it all, and that's the frame debugger...

[molten jasper]

What's the difference between square and circle previews?

It's kind of annoying not being able to use raw colour for normals

[thin crater]

Well I've recreated urp settings with the render pipeline converter, create again the same shader, created a new test shader, tested in build and nothing seems to work 😦

In the editor this shader works pretty well, but in the build or playmode it doesn't... Not even in new cameras, or with all the settings changed

What do I have to look for in the frame debugger?

The tutorial I was following also said that opaque texture had to be activated, and I did it, but nothing changed...

[jade hinge]

in the image you posted, opaque textures is turned off

[thin crater]

I know, I turned it on after that

[jade hinge]

check the settings on your camera

[thin crater]

That's now

Also tried to bypass the settings by setting both "On" in the camera

[jade hinge]

why "ultra"?

[thin crater]

Why not?

[jade hinge]

check your camera object's settings

[thin crater]

It has the default pipeline

And both forced to "On"

And I've set both settings on in all the settings, just to be sure

Maybe something's wrong with the depth fade subgraph...

[modest zealot]

Hey, I have been trying to find the answer for this for a long while. I am trying to figure out a way to utilize a shader that is exactly like the Standard shader, but that it inverts reflection probes (including inverting their box projection) so an object appears see-through.

[grand jolt]

I have a Problem
I baked a Emission map but it still does not glow

[modest zealot]

Make sure its emission level is greater than 0

[molten jasper]

@regal stag Thankyou for your help earlier. I'm liking the look of this so far

Anything you think I could improve?

[grand jolt]

Yeah I set both to higher than zero but I want it to like Illuminate glowing

Ill show an example

[grand jolt]

I want it like that

yknow what this is just a prototype to make sure the core game play works
I dont need this YET
But I would still like an anwser for later

[opaque karma]

So I have a problem hopefully I can have it fixed. On my project I have a geometric shader that creates grass, the problem is it doesn't render on my pc, but it does work on other pc's, can anyone help me? If you are wondering I have tried other shaders that are similar but they do the same thing, that they do not render in both the project window and folder.

[kind juniper]

Unless you're using HDRP, materials don't project light even if they're emissive. At least not in real time. They should light other object if you bake the lights.

A more straightforward approach would be to just add a point light source to the object.

[white pivot]

are you know how to make black to transparent?

[quasi pelican]

select alpha source/from gray scale in texture settings?

or you could select rgb black (0, 0, 0) areas and set them to alpha input

[normal shuttle]

Ok this is insane. Getting Alpha value from Color is impossible? 😄 While getting it directly from A output connector is very much ok.

Am I missing something here? Or is this a bug? I need a sanity check, pls 😄

[thin crater]

What do you mean?

Getting A from color is equal to getting Alpha value from color

[normal shuttle]

Exactly... so why does SG show different results?

The Multiply node on the top is used to extract A from the color. Does not work.

[neat hamlet]

you want split, not multiply

[normal shuttle]

While the bottom Multiply node is just used as a preview proxy. Works

[thin crater]

Maybe it's only the breview bugged

*preview

math should work

[normal shuttle]

Yep, it's a bug in the preview. Oh god, SG is so annoying.

[thin crater]

As a guy who's begun with SG, I could do nothing without it ahah

Shader code for me is dark magic

[neat hamlet]

wait thats more a limitation of the preview, not a bug, its trying to show a color texture but you set rgb to 0, theres no preview for only alpha channel

[normal shuttle]

I very much prefer Amplify Shader Editor. Using it for everything. Except for VFX Graph custom shaders, these must be made in SG only. And it is a pain 😄 Even making a basic shader is pain.

[thin crater]

so it's supposed to skip alpha?

[neat hamlet]

it just doesnt show it in previews

[normal shuttle]

But why no checkerboard?

[neat hamlet]

I dont know, ask unity 😄

does it normally do that? I don't use shader graph much, also prefer code

[normal shuttle]

No idea, I use SG only rarely as I used to write shaders manually back in built-in RP days. Now I use ASE.

But Unity is forcing me to use SG for VFX Graph now.

What an awful experience.

[molten jasper]

Yooo Joyce, I love your work

[neat hamlet]

I'm still in the built in days I guess

oh thanks nax!

[molten jasper]

It's the main reason I've been learning shaders and rendering.

[normal shuttle]

I would be too, but we have started a new project where we needed many HDRP features.

[charred crystal]

Hi there, not sure if it's specifically in shaders. But how would I go about rendering a buffer of object IDs on the screen?

[pseudo narwhal]

With the various GUI systems. For example https://docs.unity3d.com/ScriptReference/GUILayout.Label.html

[charred crystal]

I mean, translated into colors. Like id of each object on each pixel of the screen

[pseudo narwhal]

That’s just an algorithm you have to implement yourself in a shader then

[white pivot]

what mean rgb black its node?

[thin crater]

hey guys, I really hate do be pedantic but I've lost so many hours on that and I still can't find a solution T.T seems like everyone that has this problem on the internet has just depth texture disabled, and mine is activated... but it still doesn't work 😦

I've also upgraded unity in a new version, just to be sure.. nothing changed

if anyone has any suggestion.. it would be really appreciated

[neat hamlet]

have you tried turning off the pixel effect?

just to see if its the rendertexture causing the issue somehow

kinda looks like its there in your pixellized version, just mostly gone becauseo f the averaging of colors

[thin crater]

Yes, even with a freshly new camera...

Btw the pixel effect is just a camera rendering in a small scaled render image, and a canvas shows it

it should not break anything anyway

[thin crater]

It's like that even in full resolution cameras 😦

[regal stag]

Is post processing disabled on the full resolution camera? (I assume you wouldn't want to apply it twice - but this might also be turning off the CopyDepth / DepthPrepass)

Might be better to have it on, but remove all layers from the Volume Mask

[thin crater]

..it worked, seems like having more than one camera creates problems

once again, thanks

however it doesn't actually work in the same way as the scene view, seems like distance is a bit different

but i'm gonna adjust settings in a different way, so it will work anyway i thin

[hot scaffold]

Does anyone now how possible it would be to have a "fake edge" on an area with cutout?

example, a wall with an alpha circle in the middle with alpha clip enabled. this would make a circle, but make it look like it has thickness in the hole instead of just a sharp flat circle

[heavy ether]

hey, does any one know what im supposed to connect the Instanced ID node in shader graph to get value of the properties i set in DrawMeshInstanced call?

[ripe grove]

okay, so guys, this is a custom shader and this is pretty performance heavy, i was thinking if there is any way to turn this into texture map? like i dont have any interaction with it, its a static shader but it has this sort of height which gives off the depth feel, any way to turn this into texture which retains all these details?

[karmic hatch]

Put a light in the center of the cube? (you can set its parameters from a script)

[grand jolt]

Is there any difference in the building in 3D URM projects? or is it just 3D Base except with better graphics?

[regal stag]

Could use some parallax to do this

[hot scaffold]

holy, that's exactly what I want! thanks! I'll check that out

wait are you the one who did the pumpkin thing? I found that while searching for info on this!

[regal stag]

Personally I haven't used the Instance ID node, but I have used a Custom Function to handle instanced properties. https://www.cyanilux.com/faq/#sg-gpu-instancing

[regal stag]

Maybe, I did some pumpkin stuff with parallax a while ago

[hot scaffold]

yes, it was you! what a coincidence! I'm totally new to shaders, is there some way to add this to URP lit? I can't seem to find a shader graph template for urp lit or anything

[regal stag]

You'd need to recreate it. It's mostly setting up properties in the Blackboard and a few Sample Texture 2D nodes connected to the various ports on the Lit graph master stack.

In the case of the Albedo and Normal map you'd swap out the A inputs to those Lerp nodes in the graph I posted above.

Supporting Height Map is a little different. Would use Parallax Mapping node in UV ports

[hot scaffold]

thanks for all the help, very useful. Will need to learn how to use this info!

[regal stag]

@hot scaffold I should also mention that if any objects need to interact with the hole the effect kinda breaks down.
This technically could be fixed by adjusting the depth of fragment output (SV_Depth), but Shader Graph doesn't expose this (in URP at least)

[hot scaffold]

nope. it's going to be purely for visuals. will it work on geometry that isnt flat?

[hot scaffold]

that is good to know though, thanks!

[regal stag]

Eh, I think it should work. Though for very curved surfaces it might look a bit strange

[hot scaffold]

I will definitely have it a lot thinner then that. will see, thanks!

[grand jolt]

FIxed the glowing problem
I just needed to ditch the Built in renderer

[kind juniper]

Anyone ever encountered an error like that at compute shader compilation?🧐

Shader error in 'TerrainBaseGenerator': Compilation failed for kernel 'CSMain' [0x80004005 - unknown error] 'internal error: compilation aborted unexpectedly
' at kernel CSMain

Seems to be the macro...

#define Splatmap(i) Splatmap##i

Any clue what's wrong with the macro?

Ok, it's not the macro. The error is thrown even if I write down the variables manually.🤔

Ok, the problem seems to be with using an indexer to set a float4 texture2d value as float.

Like this

Splatmap0[id][channel] = value;

damn. The compiler really doesn't like me using indexers for vector types. xD

[modest zealot]

Hey, I have been trying to find the answer for this for a long while. I am trying to figure out a way to utilize a shader that is exactly like the Standard shader, but that it inverts reflection probes (including inverting their box projection) so an object appears see-through.
I feel like this should be easy, but everything I have found online has not done what I want to do and doesn't look right.

[orchid tusk]

Hey everyone. I'm using the Ellipse node in Shadr Graph. Is there a way to blur it?

[vague ginkgo]

is there any way to get the UV tiling value as a vector2? I need it for doing some vertex math but it seems to be factoring in offset as well

ass you can see the preview on the bottom isn't the same output as the vector2 on the top, despite the UV having the same tiling values

[regal stag]

No, but you can do something like one of these, if you need a gradient falloff.
Top one is closer to what the Ellipse node does, allowing for different width/height.
Bottom one is more compact if you just need a circle.

[regal stag]

These aren't the same thing.
The top is just (1, 1)
But the bottom here is equivalent to UV * (UV * 1 + 0) + 0, simplifying to UV * UV
It's a bit unclear what you are trying to do

[vague ginkgo]

I need to offset my vertex based on the tiling value of the UV

I'm scaling up my vertex output based on the texel size so I don't need to adjust my quads manually for each differently sized sprite

but if my material is only using a tiling value of .5, I need to halve my texel size(in my math)

[regal stag]

There's a Split Texture Transform node you can use on the Texture2D property

In v12 / Unity 2021.2 at least

[vague ginkgo]

awesome, thank you!

[orchid tusk]

You da man. Thanks very much!

[high hemlock]

With the shader graph, how would I go about going from local UV space to world space, as seen in material B? Thanks.

[kind juniper]

The most straightforward way is to divide and floor(or truncate) the world position by a certain "scale" value.

You'd need to then decide how to swizzle the vector3 into vector2

Using the surface normal is probably the most appropriate way

[grizzled bolt]

Triplanar node works also, but with less control

[tight phoenix]

https://github.com/yumayanagisawa/Unity-Thin-Film-Interference/blob/master/Assets/Interference.shader
Can anyone tell me what 'rd' stands for in the context of this shader? Reflected/Refracted Direction maybe?

im struggling to reproduce this shader in shadergraph because a lot of the shader is them manually calculating things that shadergraph does automatically, like sampling cubemaps or fresnel

Is there more to Thin Film Interference than just lerping a rainbow gradient across a fresnel? Am I going through all this work/steps for something that can ultimately be done much simpler?

I am not skilled enough to tell which parts I need to reproduce and which parts I don't, for example tons of this code is all about reflection and refraction and sampling colors, but shader graph to my knowledge does most of that for me, but here they have to do it manually such as setting up an SDF for a sphere, when I just have a sphere

[grizzled bolt]

As far as I understand surface viewing angle, film thickness and light bounce angle are all factors in the color effect, and to simulate it perfectly it'd be necessary to simulate photons of different wavelengths
I think most games don't bother putting more effort into it than at most sampling a fresnel and a thickness texture for a gradient as the specular tint color

[tight phoenix]

Yeah, I do like the look of their end result though. I want to reproduce their work, and then downsize to most likely a fresnel + gradient after I have a stronger understanding of the 'real world' physics version

Following a different tutorial on the same effect, they define "cosI" as 'the cosine of the incident angle, that is, cos0 = dot(view angle, normal)'
but they also have cosT but dont define what it is

I assume something to do with cosines and tangents?

https://docs.chaos.com/display/OSLShaders/Thin+Film+Shader from here.

float rs(float n1, float n2, float cosI, float cosT) {
    return (n1 * cosI - n2 * cosT) / (n1 * cosI + n2 * cosT);
}```
What would cosT be in this context?

following these people's shader work is obnoxious because every variable has some two letter name that I lack the knowledge to intuit its meaning

[leaden turret]

what's a more performant manner of making scene props have an idle animation similar to trees swaying - i dont need very fancy "wind direction" computations, even just a looped light sway (similar to idle breathing) movement would be good enough. is shaders the way to go? or something pre-baked? animations? other?

or if using dots/ecs then perhaps a system which performs the sway movement?

[karmic hatch]

cosT as cosine of angle of transmitted ray?

[shadow locust]

vertex shaders will be pretty damn good

[leaden turret]

@shadow locust oh i forgot to mention - i'm targeting mobile devices, and even hoping to support older/lower end OpenGLES 2 devices. does that affect your answer?

[shadow locust]

not really

but idk that much about low end android devices

so I might not be the best resource.

[leaden turret]

does it make sense that a vertex shader will be just as performance-intensive (or not) as a pre-made movement loop or animation or some other solution?

[low lichen]

You will need a vertex shader to render the trees anyway. An additional sine wave or two to displace the vertices will not affect performance greatly.

[leaden turret]

@low lichen can you please explain why i need a vertex shader anyway? i thought i just have a tree model and place it in the scene with a material or 2

[low lichen]

To render anything on screen, you need a vertex shader and a fragment/pixel shader. That's just how GPUs work.

[leaden turret]

so i guess you mean any shader i'm using (standard shader or other) all already have vertex shader parts to it ?

[low lichen]

Yes

And you won't be able to use the Standard shader, unless you modify its vertex shader to include the wind animation.

[leaden turret]

i see.. so even for some older Android 5 or android 6 device, chances are it would be similar performance "cost" if i make the tree sway based on some animation or other solution, or if i code it into the vertex shader?

[low lichen]

Doing it directly in the vertex shader is going to be much cheaper than a skinned mesh renderer animation.

[leaden turret]

ah ok even better then

[low lichen]

Especially if you implement the animation with some basic math and not some huge pre-baked animation

Anyway, if you're targeting low-end mobile, you should be using lower quality shaders than the Standard shader.

[leaden turret]

i see ok

[low lichen]

If you want to squeeze out as much performance as possible, it's always best to write your own that is tailored and optimized for your game and not a general purpose one like Standard.

[leaden turret]

assuming sufficient knowledge of shader & trig & GPU concepts etc....

[meager pelican]

The reason you're having a hard time is that it isn't that easy to do what you want.
I mean you want to EXACTLY match the standard shader, except for one thing where you essentially want to reverse a vector for a cubemap and that supposedly gives you "see through" effect.

This is confusing to me, as compared to making a transparent "see through" object, but OK, let's go with that.

So what you'd have to do is to clone the standard shader. They give you source. It's got a ton of includes, and it is a beast because it is an "uber" shader of sorts. If you get lucky, you can isolate everything you need to change to just one include file or something. There's a ton of variants, and I can't point you at the proper place to get to the cube map reflection calcs, but you'll find it.

I wish I had an easy answer for you, but you're in for some work IMO unless you get lucky and can just copy it over and change one include file or redefine a macro. I can't tell you the answer off the top of my head, and it would cost me hours to dig through it.

But, grain of salt, that's my 2 cents.

If you're on URP or HDRP, IDK what to tell you other than finding examples where someone did their own lighting. But I'm assuming BiRP since you discuss the "Standard Shader".

[native dome]

Will this work for converting between roughness to smoothness?

[regal stag]

You would want to use a Sample Texture 2D node, not the Gather one
But yes, One Minus should be what you need for converting between roughness and smoothness.

[native dome]

Awesome, Thank you!

[tight phoenix]

Ill try sending that in 🤔

[native dome]

Is height map supported in shader graph?

I found a 3D model online that has both a Normal map and a height map.
are they the same? I don't think Unity has a way to apply them both to a material

[karmic hatch]

You can apply a height map using parallax I think? (I haven't used it myself though)

Depends if the features are big enough that you'd notice parallax though; if they're small then you might only need a normal map.

[tight phoenix]

ignoring the color, how can I reproduce this curveature of light to dark values?

I almost sorta have it at the power node, but its backwards, I got the curvature in black and in the wrong corner

I cant figure out how to make the black part white and the white part black and also change the bottom right to top left

oh might have gotten it 🤔

[tight phoenix]

not making much headway on thin-film interference

I have a ton of really great mathematical dissertations on it, but I can't comprehend most of the mathematics within

as usual, I wont be able to make any progress on this until someone solves the hard parts for me 😿

https://www.shadertoy.com/view/ld3SDl
https://github.com/yumayanagisawa/Unity-Thin-Film-Interference/blob/master/Assets/Interference.shader
very complicated shader toy and unity shader implimentation of it, problem is they also do 500 other things and I cant separate just the part I need

[molten jasper]

mood. Definitely worth familiarising yourself with the maths imo

[tight phoenix]

I have been trying, but I don't have a university degree in mathematics, so translating this into code is basically impossible for me

not only can I not translate that into code, I can't even be sure if any of it is actually the thing I am seeking/need to solve thin film intereference

https://www.gamedev.net/tutorials/_/technical/graphics-programming-and-theory/thin-film-interference-for-computer-graphics-r2962/
sourced from here

https://openstax.org/books/university-physics-volume-3/pages/3-4-interference-in-thin-films#:~:text=Thin-film interference is most,λ n %2F 2 %2C… .

the math seems less intimidating here

[molten jasper]

https://twitter.com/freyaholmer/status/1436696408506212353

[tight phoenix]

but maybe its not showing the full story

[molten jasper]

embed you coward

[tight phoenix]

I opened the link either way

[tight phoenix]

summation of... symbols

fancy A fancy E to the power of fancy I and fancy P

brackets to the power of K

[molten jasper]

Maths likes to use single characters for variables in order to be more elegant for some reason. If it helps just swap them out for whatever they're saying it is

[tight phoenix]

so are those just A, E ,I and P? Not special constants?

like how the pi symbol is 3.14

https://sdm.scad.edu/faculty/mkesson/vsfx319/wip/best/best_winter2014/william_cavanagh/index.html

[molten jasper]

is it this one?

[tight phoenix]

another one of the many pages I have open on this topic

It doesnt super matter which one because I have no idea what any of them are calculating, helping me understand that one alone is less useful than telling me what I should be looking for 🤔

[molten jasper]

Fun fact about the use of pie to mean that irrational constant. When Archimedes was writing it he used the value to refer to any value, he just used it to mean that constant more often and people adopted it as such

[tight phoenix]

gross

[molten jasper]

In some of his works it's written "Let 'pi' be equal to..." rather than a specific thing

[tight phoenix]

Thats interesting and extremely confusing

https://www.vedantu.com/question-answer/relation-between-wavelength-and-refractive-class-10-physics-cbse-608a590bf4e14b3491a97f79

[molten jasper]

It's the same way it is today, every symbol here except infinity has a defined value

For example alpha and beta are defined above

[tight phoenix]

does anything in that link refer to thin film intereference in any way?

[tight phoenix]

I assume fancy P and fancy T are defined even higher as well?

Phi and Theta I suppose

[molten jasper]

I think that's Tau, Theta is a circle with a line thru it

[tight phoenix]

ah

I dont feel like this is useful knowledge, short of going back to university for several years, I'm not about to start to comprehend this kind of advanced math

smarter people than I have already written thin film iridescence formulas

I just have to translate their work into shadergraph

https://github.com/yumayanagisawa/Unity-Thin-Film-Interference/blob/master/Assets/Interference.shader

like its completely done here

but its also doing so many other things that it becomes useless to me

[molten jasper]

you don't need to understand the implicit meanings, choices, or workings of the formulae. Only that symbols are defined as some value or the result of some function and that it's all translatable to code

[molten jasper]

In fact the initial paper here provides an implementation at the bottom of the paper

[tight phoenix]

are you talking about the COLLOSAL block of unbroken red text?

[molten jasper]

Sure am

I blame that on the formatting

[tight phoenix]

I have zero ability to parse that yeah

i guess i can try copying it into nodepad and linebreaking every semicolon

[molten jasper]

gimme a sec

[tight phoenix]

thankfully I (think) I have already the BDSF formula

oh wait they want BRDF, this is BSDR

not that I have ANY idea what either of those things are

i google it, i found picture, i remade picture

[molten jasper]

// cosI is the cosine of the incident angle, that is, cos0 = dot(view angle, normal)
// lambda is the wavelength of the incident light (e.g. lambda = 510 for green)
float ThinFilmReflectance(float cos0, float lambda) {
    const float thickness;
    // the thin film thickness
    const float n0, n1, n2;
    // the refractive indices
    // compute the phase change term (constant)
    const float d10 = (n1 > n0) ? 0 : PI;
    const float d12 = (n1 > n2) ? 0 : PI;
    const float delta = d10 + d12;
    // now, compute cos1, the cosine of the reflected angle
    float sin1 = pow(n0 / n1, 2) * (1 - pow(cos0, 2));
    if (sin1 > 1) return 1.0f;
    // total internal reflection
    float cos1 = sqrt(1 - sin1);
    // compute cos2, the cosine of the final transmitted angle, i.e. cos(theta_2)
    // we need this angle for the Fresnel terms at the bottom interface
    float sin2 = pow(n0 / n2, 2) * (1 - pow(cos0, 2));
    if (sin2 > 1) return 1.0f;
    // total internal reflection
    float cos2 = sqrt(1 - sin2);
    // get the reflection transmission amplitude Fresnel coefficients
    float alpha_s = rs(n1, n0, cos1, cos0) * rs(n1, n2, cos1, cos2);
    // rho_10 * rho_12 (s-polarized)
    float alpha_p = rp(n1, n0, cos1, cos0) * rp(n1, n2, cos1, cos2);
    // rho_10 * rho_12 (p-polarized)
    float beta_s = ts(n0, n1, cos0, cos1) * ts(n1, n2, cos1, cos2);
    // tau_01 * tau_12 (s-polarized)
    float beta_p = tp(n0, n1, cos0, cos1) * tp(n1, n2, cos1, cos2);
    // tau_01 * tau_12 (p-polarized)
    // compute the phase term (phi)
    float phi = (2 * PI / lambda) * (2 * n1 * thickness * cos1) + delta;
    // finally, evaluate the transmitted intensity for the two possible polarizations
    float ts = pow(beta_s) / (pow(alpha_s, 2) - 2 * alpha_s * cos(phi) + 1);
    float tp = pow(beta_p) / (pow(alpha_p, 2) - 2 * alpha_p * cos(phi) + 1);
    // we need to take into account conservation of energy for transmission
    float beamRatio = (n2 * cos2) / (n0 * cos0);
    // calculate the average transmitted intensity (if you know the polarization distribution of your
    // light source, you should specify it here. if you don't, a 50%/50% average is generally used)
    float t = beamRatio * (ts + tp) / 2;
    // and finally, derive the reflected intensity
    return 1 - t;
}```

[tight phoenix]

oh when its formatted nicely its much more clear, they even commented it

[tight phoenix]

is that the same as the above?

except written in maths?

[molten jasper]

I believe so

[tight phoenix]

I see 🤔

[molten jasper]

You can sort of make these out from the code directly

[tight phoenix]

Ill try to repro the above code now.
I also had this link where they put the code fairly sensibly, but I wasnt able to deduce what all the random letter variables were supposed to be, like cosT
https://docs.chaos.com/display/OSLShaders/Thin+Film+Shader

[molten jasper]

Yeah

[molten jasper]

The wave amplitude calculations you'll need to write yourself. Just implement them as called following the middle image here

[tight phoenix]

that explains why some of the shaders I looked at were calling things 'rs'

people copying people copying people

[molten jasper]

r = reflection,
t = transmission
s = s-polarised
p = p-polarised

[tight phoenix]

now to google all of those things

[molten jasper]

https://en.wikipedia.org/wiki/Fresnel_equations?oldformat=true#S_and_P_polarizations

[tight phoenix]

how much of all these incredibly deep mathematics are being hand-waved/solved by this node by itself

[molten jasper]

it gets a bit physics heavy here on in lol

[tight phoenix]

yeah 🤔

[molten jasper]

An incredible amount

[tight phoenix]

I wish I could tell which parts I can safely drop/subtitute for pre-existing nodes

[tight phoenix]

this is getting pushed up more and more, starting on implimenting it now before its lost

[molten jasper]

want me to DM it you so it doesn't get lost?

[tight phoenix]

Im saving it into a .shader file in my project now

// cosI is the cosine of the incident angle, that is, cos0 = dot(view angle, normal)

that part is the 'fresnel' bit, if I did that right?

[molten jasper]

https://www.youtube.com/watch?v=gfK8pVf_dww

[molten jasper]

this isn't important to know for this but I got curious since wikipedia's article demands a certain level of prerequisite knowledge to learn

[molten jasper]

I think it's meant to the cos Theta but they chose 0 for some reason

[tight phoenix]

this is all branching logic, can that even be done in shader?

unless its not branching logic at all

my translation of that is "if n1 larger than n0, true = 0, false = Pi

[molten jasper]

that's correct

you can do logical branching in shader graph, I've not touched it myself however

[tight phoenix]

googled it

this feels dirty, everything ive ever been taught about shaders is no branching logic

[molten jasper]

The problem with conditionals is branch prediction. That's why tricks like multiplying is preferred

[tight phoenix]

thankfully it converges pretty fast

[molten jasper]

btw they're set as constant floats with no initial values there but the paper defines them as follows:

Consider a light wave incident to a thin layer of depth \deltaδ and real refractive index n_1n 1. The external medium has refractive index n_0n 0 and the internal medium has refractive index n_2n 2. The incident angle made by the incident light wave and the film's surface normal is \theta_0θ 0, the angle inside the layer is \theta_1θ 1 and the refracted angle (inside the internal medium) is denoted \theta_2θ 2 . We will also give numbers to each of the three media: medium 0 is the external medium, medium 1 is the layer, and medium 2 is the internal medium. Also, naturally, medium 0 has to have a different refractive index than medium 1, and the same goes for medium 1 and medium 2. Media 0 and 2 can be the same, of course.```

First paragraph under the heading derivation

This site is terrible at formatting

[tight phoenix]

sooo they're saying n0,1,2 are values of 0, 2, and theta?

yeah the formatting makes it nearly unreadable

[molten jasper]

I think layer one is the thin film thickness? Near the end they state if you set that to 0 it plays out like regular fresnel

[tight phoenix]

I feel in my heart at the end of this the end result is going to be 'throw all of this out and just use fresnel on a color gradient 🔥

but its the learning that matters, not the result

[molten jasper]

sure but bubble dragon

[tight phoenix]

hell yeah bubble dragon

okay contining onto the next nodes

at least when I am done it will be easy to tell what can be replaced for what

if (sin1 > 1) return 1.0f;
Saturate and Clamp both do values 0 to 1 right?

Normalize rescales its values to be between 0 and 1, but saturate is what they're doing here?

[molten jasper]

Normalise scales to 1

[tight phoenix]

    // rho_10 * rho_12 (s-polarized)
    float alpha_p = rp(n1, n0, cos1, cos0) * rp(n1, n2, cos1, cos2);
    // rho_10 * rho_12 (p-polarized)
    float beta_s = ts(n0, n1, cos0, cos1) * ts(n1, n2, cos1, cos2);
    // tau_01 * tau_12 (s-polarized)
    float beta_p = tp(n0, n1, cos0, cos1) * tp(n1, n2, cos1, cos2);````
Ive gotten this far now

all four of those are functions right?

im guessing those functions maybe are elsewhere on that page?

[molten jasper]

yup as defined here

[tight phoenix]

😰

rs = n1 cos theta lower case i minus n0, cos theta lower case j

divided by.. the exact same values?

oh but plus instead of minus

Im going to have to make a subgraph for this

[tight phoenix]

uhh there's no symbol between Ni and cos0

is it multiplying?

[molten jasper]

yep

[tight phoenix]

is I some kind of function? oh okay

[molten jasper]

no operator between values means multiply

[tight phoenix]

I guess I knew that somehow but couldnt remember why

[molten jasper]

btw i think you can drop the conditionals and subtract and pull the sign

[tight phoenix]

oh there IS a pi node

[tight phoenix]

Ill update mine to that

[tight phoenix]

So you are saying these are equivalent?

[molten jasper]

I believe so

[tight phoenix]

Ill keep the original off to the side just in case

[molten jasper]

if n1 is greater than n0 then the result of subtracting them should result in a positive number, which sign returns as 1, otherwise it should result in a negative number, giving us -1

on that note I'm adding a max after the result of sign with 0 as B since we want to use that multiplier to determine whether we want 0 or pi

[cursive anchor]

where can i find this setting

[molten jasper]

then the result of that into the multiply

[tight phoenix]

its starting to get messy and involve a lot of subgraphs but progress is being made

[molten jasper]

ikr lol

[tight phoenix]

oh god I just saw how many times these polarizing function things happen

eight of them

I can no longer keep track of it and im 100% sure im making a fuckload of mistakes

I know im not going to be able to finish thsi tonight

im trying to carefully re-create this

but I think its too complicated

when this doesnt work because ive made tones of mistakes, its going to be hell to find them all

[molten jasper]

screw it I say

[tight phoenix]

Thats a good idea

[molten jasper]

_out = (ni*cosi-nj*cosj)/(ni*cosi+nj*cosj);
_out = (2*ni*cosi)/(ni*cosi+nj*cosj);
_out = (nj*cosi-ni*cosj)/(ni*cosj+nj*cosi);
_out = (2*nj*cosj)/(ni*cosj+nj*cosi);```

for rs, ts, rp, and tp respectively

[tight phoenix]

thanks, I just finished typing the first one from your text, copying will be faster

[olive aspen]

I need to know what that shader is right now

[tight phoenix]

its the one im trying to write as we speak

[molten jasper]

https://www.gamedev.net/tutorials/_/technical/graphics-programming-and-theory/thin-film-interference-for-computer-graphics-r2962/

[tight phoenix]

Thin Film Interference

[olive aspen]

Thank you good sirs

[tight phoenix]

if you finish it before I do, pls share your work because its far from easy 🔥

I will do the same of course

[olive aspen]

oh no I will just see skim the article, I am far from this level

[tight phoenix]

these websites have assisted me this far

rofl fair, so am I!

[olive aspen]

good luck!

[molten jasper]

ty

praise be the mighty tower

[tight phoenix]

totally not spahgetti

float phi = (2 * PI / lambda) * (2 * n1 * thickness * cos1) + delta;
// lambda is the wavelength of the incident light (e.g. lambda = 510 for green)

how do I know what the light's wavelength is?

I have the main light's direction

[molten jasper]

I honestly have no idea, for now I'm setting it as a property

[tight phoenix]

I will do the same

I will be amazed if this does anything at the end

right now it looks like nothing but tons of white circles

[molten jasper]

same lol

[tight phoenix]

    float tp = pow(beta_p) / (pow(alpha_p, 2) - 2 * alpha_p * cos(phi) + 1);```

to the power of what?

[molten jasper]

fun

Not sure, given the right hand of the divide I'm going to assume 2

[tight phoenix]

two is the default in shader graph as well

god I hate assumed info

[molten jasper]

Sounds good to me

[tight phoenix]

like no symbol means assume multiply

[molten jasper]

I've been using mul instead of pow for those

and just passing the same value in both A and B since n*n == n^2

[tight phoenix]

I see 🤔

wait hold on

the next one says 2 yheah

I guess its both two???? but it feels really wrong to just assume that

[molten jasper]

it is rather annoying

[tight phoenix]

making up stuff that isnt in the source is how you get end results that dont work

Ill make a note of this in case its not working to check here first

float ts = pow(beta_s) / (pow(alpha_s, 2) - 2 * alpha_s * cos(phi) + 1);

im pretty sure im fucking up the PEMDAS of this

I knwo you do all mults/divides before addition

but im not sure where the addition goes after the mults/divides resolve

also is there missing brackets?

[molten jasper]

nah it's BIDMAS

[tight phoenix]

because there's a ) at the end, but it has no start

[molten jasper]

so resolve priority first then LTR

oh I misread that

[tight phoenix]

Ill have to google both bidmas and ltr

when I select this final bracket, it has no opening bracket

im hoping this is just a mistake on their part

[molten jasper]

it's the bracket before the second pow

[tight phoenix]

my nodes are all broken purple

not a good sign

[molten jasper]

bad omen

oh your n0/1/2 are all 0 right?

div by 0 err

[tight phoenix]

my Ns are not zero

allegedly this is done, now I just have to trace the purple

alpha P is zero

okay whty is alpha p zero becomes the question

alpha S is also zero

are yours zero?

what values are you using?

I have no idea, no ability what so ever, to diagnose what is wrong here

its just too big

and too much mess

these things return zero but I dont know why they return zero or how to fix it

my anxiety is spiking

that I did all this work for nothing

my mistake was hoping that if I just copied it exactly that it would work

but that was not a realistic goal

nothing ever works when you copy it exactly becaue I'm a stupid human being who didnt copy it excvtly

[molten jasper]

Mine isn't working either, dwai

[tight phoenix]

i just have to redo the entire thing from scratch step by step

until I find the mistake

these things are all returning zero

the mistake must be in here

[tight phoenix]

maybe this is NOT representitive of the text code block???

[molten jasper]

that could be wrong, I did it rather fast and didn't double check everything

[tight phoenix]

ill ready it very very carefully

[molten jasper]

especially given both i and j look exactly the same in their dog shit font

[tight phoenix]

yeah

do we even know what Ni and Nj are

like which n0, n1, n2 is nj or ni

[molten jasper]

n is referring to the nth index and i/j are referring to the set of parameters

[tight phoenix]

oh I found a mistake

https://cdn.discordapp.com/attachments/497874081329184799/1015014677933203517/unknown.png

ts and tp have a different formula

but in yours its all the same formula in all 4

oh wait no

its me

the mistake is that I didnt copy the right things into the right custom nodes

ill fix that now

well its still purple at the end but at least that part is now showing something

[molten jasper]

[tight phoenix]

hmm problem appears to be only in beta S and alpha s

[tight phoenix]

at least its not purple!

these are all returning zero

but its identical to the code

float alpha_s = rs(n1, n0, cos1, cos0) * rs(n1, n2, cos1, cos2);

this is that right???

and RS is Out = (ni*cosi-nj*cosj)/(ni*cosi+nj*cosj);

[molten jasper]

that's what I have yeah

[tight phoenix]

does your return zero though?

this area is the problem area

im positive i pemdas'd this wrong

thats the only thing I can see that could be causin this

that I can see

with my stupid inexperienced eyes that see nothing

[molten jasper]

lol

[tight phoenix]

😱

my beta s is zero, this is a problem

maybe its a values problem

might be something to do with the values

maybe it needs very specific values to work???

[molten jasper]

I have mine set to 0.1, 0.2, 0.1

[tight phoenix]

accidentally accessed the chaos dimension

floating point nightmares

[molten jasper]

praise be the eldritch lords that preside over precision

[tight phoenix]

im going to go ahead and say i have no idea whats wrong, and no idea how to fix it

and probably i should just cut my losses and move onto to adampting another person's version

https://sdm.scad.edu/faculty/mkesson/vsfx319/wip/best/best_winter2014/william_cavanagh/index.html

like this persons's

this oen actually talks about RGB colors so that already feels more like it might work???

I think this guy is doing the exact same math

[molten jasper]

oh I missed something

[tight phoenix]

because he's got n0, 1, 2, he calls cos0 eta0, and delta is there too

[tight phoenix]

Oh? 👀

[molten jasper]

We can now sample this function at red, green, and blue wavelengths (650, 510, 475 nanometers, respectively) and substitute the RGB reflectance obtained into the Fresnel term of the BRDF

[tight phoenix]

yeah this is definitely the same math, he has delta10 and delta12

oh dear god are they saying to run that gigantic mess three times

[molten jasper]

yup

[tight phoenix]

this is starting to look more and more appealing

[molten jasper]

lol

[tight phoenix]

albiet completely chromatically incorrect

I do have this from a totally different tutorial