archived-shaders

[tight phoenix]

I think my problem is that voronoi cells get distance from their center, but I need distance from the edges?

because the edges are not all uniformly the same value, they're all over the place, its only the centers the same value, so I am not able to mask the edges

https://www.shadertoy.com/view/MtlyR8

but like, this person 100% is getting the edges

but I can't seem to parse their code

[regal stag]

The "Out" preview on the NewVoronoi node looks like distance to edges, but also looks like regular voronoi not the Chebyshev one

[fluid salmon]

im using a blend node in overwrite mode to overlay an image with transparency (overlayImg) onto another texture. How can I make it so that this step doesnt happen if overlayImg is empty? At the moment, if it's empty, the overwrite turns the output fully white.

Would this be a branch node or something?

[tight phoenix]

https://paste.ofcode.org/kEQJVC68wXkX5hKW5iMVYV this is the code of the most recent attempt, I've started over from scratch every time I hit a wall I can't solve on my own so I am so neck deep in variations I no longer can see what any of them do

This only makes this harder, my comprehension sucks no matter how many hours I put into it, it feels like nothing ever becomes clear

[regal stag]

I assume you're using the alpha channel of the texture as the blend amount? If you change the Texture2D property to "Black" under the Node Settings (rather than White) it should default to 0 alpha.

[fluid salmon]

I was! Thank you that worked

[regal stag]

I have a feeling these sort of calculations might be assuming an Euclidean distance still so might be why it doesn't give the correct output.

float d = dot(0.5 * (mv + v), normalize(v - mv));
res.x = min(res.x, d);

I'll play around a bit and see if I can figure it out

[tight phoenix]

currently googling more shadertoys looking for one that has less superflous features in it, its always so hard to parse what is doing the thing I want from the mash of extra visual flair stuff

[tight phoenix]

Alright, Ill keep poking on my end as well. There are some distance functions for various distance spaces at the top of the code if those help / are useful / are even correct

[regal stag]

Chebyshev Voronoi Edges

[tight phoenix]

my scene looks dramatically different in build compared to game view of the editor

how can I debug whats occuring here?

I cant click on and inspect any of the assets in the build version, and those are the ones I need info about

I have the debug profiler hooked

but how do I find that sepcific thing?

[tight phoenix]

I found the specific thing 🤔 through manual experimentation with what I figured it might be

Not sure how to fix it yet though

the problem is that I'm running this: ``` matPropBlock.SetFloat("_GlobalBrightness", value);

that value is getting set to 0 in build but not in editor

I dont know how that is possible because nothing in the scene is even scripted to do that

and its default value is 1

do MaterialPropertyBlock work differently between editor and build?

[tight phoenix]

ah, found the deeper actual problem

I am writing barycentric coordinate data to triangles, but the method that does this can't run at build runtime

the shader needs that data to be visible, thus why it was vanishing

[raven hamlet]

Can someone help me understand why this code isn't getting me a fragment's OS position?

                // clipspace frag position
                float u = IN.vertex.x * 2.0 / _ScreenParams.x - 1.0;
                float v = IN.vertex.y * 2.0 / _ScreenParams.y - 1.0;
                float4 ro = float4(u, v, -1, 1); 

                // obj space frag position
                ro = mul(UNITY_MATRIX_IT_MV, mul(unity_CameraInvProjection, ro));

[frigid jay]

Object space position is just what you get in vertex shader in parameter with SV_Position (or POSITION) semantic. Your code trying to (incorrectly) make object space position of pixel that reside behind camera at imaginary far plane.

[raven hamlet]

SV_Position is screen position right? e.g. in 1920x1080 space yeah?

[frigid jay]

No. I'll give you an example

[raven hamlet]

"... the SV_Position semantic (when used in the context of a pixel shader) specifies screen space coordinates (offset by 0.5)"

[frigid jay]

struct v2p
{
  float4 pos: POSITION;
  float3 objectSpacePos: TEXCOORD0;
};

v2p vs(float3 vertex: POSITION)
{
  v2p o;
  o.pos = UnityObjectToClipPos(vertex);
  o.objectSpacePos = vertex;
}

float4 ps(v2p i): COLOR0
{
  //  i.objectSpacePos is your object space position in pixel shader
  return float4(i.objectSpacePos, 1);
}

Something like that

[frigid jay]

This is correct

[raven hamlet]

ah ok

So the space we're in in the fragment shader is clipspace

I'm not sure how unity handles the nearplane, I've tried both 0 and -1 which afaik are the standard options

[near fable]

Hi, I just imported an asset I got from the asset store and the materials are all pink. I had a look and they're using custom shaders. I'm also getting this issue:

UnityEngine.GUIUtility:ProcessEvent (int,intptr,bool&)```
Unity version 2021.2.8f1, and using URP

[raven hamlet]

I'm trying to get the position of of the fragment in object space for the origin of a raymarching ray. I have the position of the object in object space (the hit of the ray) no problem

[junior iris]

Why is the UV Coordinate node in Shader Graph a Vec4? What do the ZW channels represent?

I thought UV Coords are just Vec2?

[safe ravine]

I noticed that some packages show which rendering pipelines are supported. But it is not shown in the web page for my asset. How can I fix this?

[frigid leaf]

Hey everyone. I have a standard set of shader VFX I need to apply to all my models (like glow, fresnel, changing its albedo color). Every model will have the same pack of effects. I remember that it's generally not recommended to put multiple materials into a single mesh. Also, our game will have mods and that probably will include adding user VFX. How do I do this right? Do I have to combine both my surface shader and VFX into a single material or can I split them into multiple separate materials? What are drawbacks? (I'm using HDRP)

[fossil cedar]

this seems more like an asset render pipeline compatibility question than a shader question. either contact the publisher of the asset for support (if they support URP,) or find another asset that is compatible with URP.

[fossil cedar]

not really a shader topic, this is an asset publisher portal question. i suggest to try another channel like #💻┃unity-talk , the forums, the docs, or contact unity asset publisher support.

[fossil cedar]

the zw components could be used for a second set of uv coordinates, tiling and offset for texture atlases, sampling cube maps or 3d textures or a completely novel technique that no one has ever thought of. anything really ¯_(ツ)_/¯

[fossil cedar]

it sounds like you may want to look into HDRP custom passes https://docs.unity3d.com/Packages/com.unity.render-pipelines.high-definition@16.0/manual/Custom-Pass.html

[fossil cedar]

fortunately, there's a number of raymarching tutorials and examples out there for unity. i recommend to follow one of these or at least examine their code as a frame of reference https://youtube.com/playlist?list=PL3POsQzaCw53iK_EhOYR39h1J9Lvg-m-g
https://youtu.be/S8AWd66hoCo

[raven hamlet]

I got it figured out

[raven hamlet]

transposing the matrix multiplication to go from camera space to object space ended up fixing the issue

                float4 ro = float4(u, v, -1, 1);
                float4 camSpace = mul(unity_CameraInvProjection, ro);

                // obj space frag position;
                ro = mul(camSpace, UNITY_MATRIX_IT_MV);

[smoky widget]

Is there a way I can make a camera render red and green in a normal way, but get only the minimum of the blue channel?

Like from all the colors that are affecting a certain pixel, get the minimum of all

[near fable]

Ah ok, thanks!

[shadow locust]

I think you would need custom shaders and/or postprocessing for this

if all your objects use a shader that takes the scene color and mins it with its own blue channel for example

[frigid jay]

Alpha blending cannot be configured per-channel. So you will need multipass approach. Set BlendOp to “Min”, Blend to “One One”, and ColorMask to “B”. Then render all interested meshes. At the end you will get minimum in blue channel.

For “RG” channels you will need separate pass. You can use alpha channel as imaginary “B” channel. Alpha channel blending can be configured separately from RBG

[frigid jay]

This approach can be easily done with RWTexture2D usage.

[smoky widget]

Where would I do this? In like a material shader used for blip onRender? Or on the material of the object which this rule applies?

If its not too much, could you write somd pseudo code for how it would look like?

I think i understand the B channel thing, but not the RG

[rare wren]

Got an issue with shadow attifacting with custom spot light shadows.
If someone can help it is greatly appreciated:
https://forum.unity.com/threads/artifacts-at-the-edges-of-spot-lights.1463939/

[smoky widget]

It's for a particle, but I cannot get it to work, by just using the unlit shader im already getting Z culling errors

[frigid jay]

Alpha blending is configured per-material

[smoky widget]

Okay, so for example, how can I get the Youngest in front sort mode of particles to work on a custom shader¿

[frigid jay]

For RG channels you should use ordinary shader just with ColorMask RG.

[frigid jay]

What kind of Z culling errors do you have?

[frigid jay]

Sorry, I don’t understand your question. Can you clarify please?

[smoky widget]

This using a default unlit shader

And this is using the default particle shader

[smoky widget]

Yes, so on my particle System, I have this enum set that makes it that the youngest is always rendered on top.

Now, what I understand is that, i need to create a custom shader that will have first a pass that blends and gets the minimum of blue, then a second pass that renders Red and green and somehow get the blue from previous pass, (not super sure how) and then defined the remaining fields that would be needed to keep that particle working too

This is as far as I got

[frigid jay]

So, you need two-pass shader. First pass will calculate minimum in R channel. Options for this pass: “ColorMask B”, “BlendOp Min”, “Blend One One”
Second pass will have just one additional property: “ColorMask RG”

[smoky widget]

Okay, so something like this

Ah! Disabling the Zwrite fixes that problem i had with the sorting!

Not sure how to remove the yellowish thing

Aha! changing it so that the first pass only returns the blue channel removes the yellow

Ah, not that's what I thought, but not really removing it

that's because blue also needs to be transparent with alpha

is that possible?

[frigid jay]

So transparent or minimum? You can premultiply color with alpha in shader

[smoky widget]

I need the minimum of all blue channels, but it's creating blue values where they don't exist (where it's transparent)

[frigid jay]

I am not sure that it what you need, but alpha premultiply can help you: outColor.b *= outColor.a;

[smoky widget]

Ah, multiplying wasn't working because where the texture was transparent i was getting a value of zero on the blue channel, where I actually needed a 1 there, the solution was

return float4(0,0,col.b+(1-col.a),0);

Now the result of all particles becomes this texture, which is what I wanted

[karmic hatch]

Maybe the scaling is different, it should be identical if world space and object space are the same (i.e. no rotation, no scaling, no offset)

[clear veldt]

Some objects have rotation, I am trying to make it work with those because they float around

Idk how to, I’ve been stuck for a little bit

[mighty fiber]

hey y'all, i'm having a little problem with some assets i downloaded. I'm working in Unity 5.6 because i'm modding a game but i tried importing assets made for unity 2020 after using the built in converter to make them usable with legacy shaders. However, something about the shaders or the code isn't working and is producing this conondrum. Please let me know if you have any suggestions.

[vale salmon]

made a shader graph called "scanned"

tried to make a mateiral from it but its just blank on the right side

tthere are no options for me to change anything like i have made the variables in the actually shader you can see on the left

the base color and fresnel power.

but i have no options for those in the inspector after creating a material from the shader graph

nvm i fixed it

didnt save asset

hm, but still have some problems. the shader isn't working as intended

the main preview is completely blank, and even after chaning settings in the material i create from the shader, nothing changes

nvm fixed

[idle copper]

Hello, is there a way to set the normal of a pixel being rendered in a surface shader? I want all pixels to be rendered with the same normal (therefore being lit the same way) regardless of the mesh. I also need shadows so I can't use an unlit material.

I did try setting the normal of the SurfaceOutputStandard, but that doesn't seem to work. Half of the mesh is rendered (kind of) normally while some faces are all black.

[shadow locust]

Could you try also setting htese to 0?

    half Smoothness;    // 0=rough, 1=smooth```

[idle copper]

I already do. Setting those to zero just gives diffuse lighting. I want the whole mesh to be lit the same even if the face is opposite the light source.

It would basically be like an unlit material that casts and receives shadows

[shadow locust]

right I wasn't sure if you were already, because I think diffuse lighting would be required for that effect

but with faces that are not facing the light can't he object itself be casting a shadow on itself?

[idle copper]

Yeah, that's why I want to set the normal of every pixel or fragment or whatever to straight up so that it uses diffuse lighting but is completely lit

[idle copper]

In reality, yes. In what I want to do, no.

Do objects cast shadows on themselves in Unity?

[shadow locust]

yeah

i think so

[idle copper]

Is there a way to disable that?

[idle copper]

Nvm, I found a shader online.

[shadow locust]

the concept is sound. I've seen it done before

maybe just an issue with your implementation

[lyric tiger]

Found this article https://danielilett.com/2021-11-06-tut5-19-glitter/ problem is now fixed

[karmic hatch]

Yeah if you use everything in world space it doesn't care about rotation vs if you use object space, it does. Only difference might be the actual numerical factor you put in for the world bending effect

[frigid jay]

MaterialPropertyBlock did not contain SetInt at that time. You need to replace ints with floats in your shader, and SetInt calls with SetFloat.

[green peak]

  Shader "Example/TestShader" 
  { 
    SubShader {
      Tags { "RenderType" = "Opaque" }
      
      CGPROGRAM
      #pragma surface surf Lambert
      
      struct Input {
          float4 color : COLOR;
      };
      
      void surf (Input IN, inout SurfaceOutput o) {
          o.Albedo = 1;
      }
      
      ENDCG
    }
    Fallback "Diffuse"
  }

why is it pink... I pulled this straight from unity

Figured it out, it needed a default Cull front

[green peak]

Nvm, it doest work

  Shader "Example/TestShader" 
  { 
    SubShader {
      Tags { "RenderType" = "Opaque" }
      
      CGPROGRAM
      
      Cull Front // this line is suppose to be before CGPROGRAM

      #pragma surface surf Lambert

      struct Input {
          float4 color : COLOR;
      };
      
      void surf (Input IN, inout SurfaceOutput o) {
          o.Albedo = 1;
      }
      
      ENDCG
    }
    Fallback "Diffuse"
  }

when I do this, the sprite works

although the code is wrong

  { 
    SubShader {
      Tags { "RenderType" = "Opaque" }
      
      Cull Front // this line is suppose to be before CGPROGRAM
      
      CGPROGRAM
      

      #pragma surface surf Lambert

      struct Input {
          float4 color : COLOR;
      };
      
      void surf (Input IN, inout SurfaceOutput o) {
          o.Albedo = 1;
      }
      
      ENDCG
    }
    Fallback "Diffuse"
  }

If i do this, the shader is back to purple

I have tested this with both Cull Front/Back/Off

what is GOING ON!!

[regal stag]

What render pipeline are you using? Surface shaders are only supported in Built-in, not URP/HDRP.

[green peak]

O

[green peak]

Im in URP

[regal stag]

Then yeah, not supported. That's why it's magenta. It's usually recommened to use Shader Graph instead

[green peak]

I figured shader graph would be limiting in the long run...is that the case?

[regal stag]

Hmm perhaps? There are some things shader graph can't do. Depends what you need the shader to do.
Can generate code from a graph (though that tends to be pretty unreadable). Could also copy one of the shaders under the URP package as a basis then edit if you really want a code one.
But it is easier with a graph and less likely for the shader to break with updates to the pipeline.

[green peak]

I see

[neat horizon]

Why is it taking so long, I don't have a shader that I created myself, I use it from the assets that everyone uses and not much

[green peak]

is ur code running?

is it paused at a break point?

post shader

you can click on material and click edit to see shader code

[mighty fiber]

Thanks!

[neat horizon]

I don't have shader code, I don't know if it's because of URP's own codes, I haven't dealt with shaders in the project

[green peak]

the assets themselves should have shader code

[green peak]

click it and send us code

[neat horizon]

There are 20 shaders in the project. I don't know which one is wrong and it doesn't show. If it gives an error after the build is finished, I will post it here

[green peak]

hmm

there is no point waiting

just restart, if it didnt work in 3 hours sure as hell wont work in 6

even if it does take 6 hours to work

its no sensible because everytime you make changes are you really gonna wait 6 hours?

[neat horizon]

ok how do i find it now

[green peak]

find what?

[neat horizon]

why is it taking so long regarding shaders?

[green peak]

could be anything really

No one understands the nature of ur program

[neat horizon]

error

public static void BlitTexture(CommandBuffer cmd, RTHandle source, Vector4 scaleBias, Material material, int pass)
        {
            s_PropertyBlock.SetVector(BlitShaderIDs._BlitScaleBias, scaleBias);
            s_PropertyBlock.SetTexture(BlitShaderIDs._BlitTexture, source);
            DrawTriangle(cmd, material, pass);
        }

this is the error code
s_PropertyBlock.SetTexture(BlitShaderIDs._BlitTexture, source);

[clear veldt]

So what am I supposed to do then? I’ve pretty much tried everything

[green peak]

any1 know why my materials arent matching?

 struct appdata {
                float4 vertex : POSITION;
                float4 color : COLOR; // Assuming the colors are provided as float4 in the C# script
                float2 uv : TEXCOORD0;
            };

            struct v2f {
                float4 pos : SV_POSITION;
                float4 color : COLOR;
                float2 uv : TEXCOORD0;
            };

            v2f vert(appdata v) {
                v2f o;
                o.pos = UnityObjectToClipPos(v.vertex);
                o.color = v.color;
                o.uv = v.uv;
                return o;
            }

            // Fragment shader
            // This shader simply passes through the color to the output

            fixed4 frag(v2f i) : SV_Target {
            
                fixed4 col = tex2D(_MainTex, i.uv);
               // col *= _Color;
                return col;
            }

[karmic hatch]

I think if you have something that roughly does

worldPos = worldPos - vec3(0,1,0)*(worldPos.x * worldPos.x / x_radius + worldPos.z * worldPos.z / z_radius)

where worldPos is the world-space vertex position, then it should create the desired effect. (If the object has triangles about as large or larger than min(x_radius, z_radius), you will have issues). There should not be any dependence on rotation or scaling with this. (You can switch the Vertex node to take a world-space position so you don't need to do any transformations yourself)

[clear veldt]

I may have a script for something like this, I’ll try it

[karmic hatch]

Are you sure the UVs on the mesh are set up correctly?

[green peak]

Not sure, still learning 🙂

[green peak]

how can I verify

[karmic hatch]

If you do col = fixed4(uv, 0, 0); it should show the UVs and it should be black in one corner, yellow in the opposite corner, and then green and red in the two remaining corners

[green peak]

Im guessing this si bad

[green peak]

Strangely enough the original mesh is fine

I am creating a procedural mesh

[karmic hatch]

yes so you're just sampling one corner of the texture across the whole mesh, which is why it's just solid color

[karmic hatch]

I believe there's a mesh.uv array which you can set

[green peak]

what do I set it too?

[karmic hatch]

if you don't want the texture stretched and it's aligned perpendicular to z, we can define d = max((max(vertices.x) - min(vertices.x)), (max(vertices.y) - min(vertices.y))) as the size of your mesh, and then you want your uvs to be Vector2((vertex.x - min(vertices.x))/d, (vertex.y - min(vertices.y))/d) where vertex is the particular vertex for that element in the array. (note that max(vertices.y) is not valid, if you don't a priori know the maximum size of your mesh, you will have to loop over the array. You probably do know how big the mesh is though since you're making it with code so it should be fine)

[green peak]

What is the best way to add depth to my meshes

Using different materials has its limits

[grave vortex]

How can I use Nsight's shader profiler with Unity?
I can't seem to get a build to include shader symbols

[clear veldt]

Wait actually how do I make a script for this or add the right nodes in the shader?

[neat horizon]

Shader error in 'Universal Render Pipeline/Lit': maximum ps_4_0 sampler register index (16) exceeded at INVALID_UTF8_STRING(9) (on d3d11)

I am getting this error after getting build

[clear veldt]

I made a shader graph that curves world but objects that have rotation have this weird offset and float around. I still haven't been able to fix this after a few days

[reef turret]

Hey, is there any possible way to use a graphics shader with a compute shader? I'm trying to use the Raymarching technique to blend meshes together to simulate a fluid like effect, I'm using URP

[lucid geyser]

Does anyone know of a great source for making unity shaders? I want to create 2d effects on a cimple circled. is something like this possible in unity ? https://www.youtube.com/watch?v=f4s1h2YETNY

like can this be applied to a basic white circle sprite?

@karmic hatch do you know of any good sources for this ^^^

[fossil cedar]

there's a few video tutorials for converting shadertoy shaders to unity built in render pipeline / shaderlab: https://www.youtube.com/results?search_query=convert+shadertoy+to+unity

[fossil cedar]

you can also look up general GLSL to HLSL resources like this which are also helpful. the languages are very similar in structure. https://learn.microsoft.com/en-us/windows/uwp/gaming/glsl-to-hlsl-reference

[fossil cedar]

or if you are using Universal RP or High Def RP, there are some converters on the asset store from this developer that would give you a head start. also for Shaderlab / Built in RP looks like https://assetstore.unity.com/publishers/72922

[fossil cedar]

yes, you can execute compute shaders that write to a graphicsbuffer or computebuffer or texture, and then read those buffers or textures in traditional "graphics" shaders i.e. vertex or pixel shaders. all told this can be quite an endeavor but there are some examples out there. here is one with links in the video's description to example code and also some assets that are further developed and more efficient https://www.youtube.com/watch?v=eisCx-pC0oo

[reef turret]

thx

[fossil cedar]

np. there's also screenspace fluid rendering, and isosurface mesh generation like marching cubes, dual contouring, or surface nets. just to give you some other techniques to look into.

[lucid geyser]

Why is my sprite showing up as this color? what

the green should be transparent and the yellow should be white

[fossil cedar]

what does the source texture look like in the inspector? (not in shadergraph, before that)

[lucid geyser]

it looks how i described

[fossil cedar]

can you screenshot that, and the texture format listed below the inspector preview

there's weird texture formats sometimes

[lucid geyser]

it's doing it with every sprite that i use

it's doing it with this sprite

[fossil cedar]

even the built in sprites?

[kindred cradle]

I've now followed two separate tutorials for some water shaders and both times after applying the shader to an object my unity editor Scene tab is very choppy.
The game runs fine and if I RMB and move around in the Scene it runs smooth.
Is this a normal issue?

[lucid geyser]

yes ^

this is a built in circle sprite

[fossil cedar]

wild, is it only in shadergraph? or does it look yellow and green in the game and scene view also?

[lucid geyser]

in scene as well

i got the transparent part to disapear but the white is yellow

this is scene^

[fossil cedar]

not exactly sure but unless i'm missing something, i feel like it could either be a unity bug, or a bug or error with your gpu or the gpu drivers

it even happens when you create a new shadergraph?

[lucid geyser]

is it possible that i'm using the wrong shader ? i'm totally new to shader stuff. just doing tuts

I also changed my pipeline asset to a custom 2D URP

here are the pipleine assets

[fossil cedar]

maybe sprite textures are weird, i haven't used them with shadergraph

i know they aren't regular textures exactly

that's why the texture importer has an option for sprites, because they are unique in some way

[lucid geyser]

I'm also new to game development so... ya i probably effed something up somewhere.. what would you double check first?

instead of using a sprite... what would you use?

[fossil cedar]

this is a few years old but there might be a special shader graph type for sprites https://kylewbanks.com/blog/sprite-shader-effects-with-unity-and-shader-graph-part-1-getting-started

[lucid geyser]

i literally just want to make effects using a circlular png or sprite whatever

i don't need to use sprites

[clear veldt]

Im in a 3d project
Sprite is just a nickname for the Texture2D

[fossil cedar]

but yeah you could also not use what unity considers sprites, just a regular 2d texture on a quad / rectangle which is what i usually do but i work mostly in 3d

[fossil cedar]

yes and no, in general game dev terms yes, in unity also yes but also a Sprite is a special "type" of Texture, that's why this is selected... as to how it's special i'm not sure

[clear veldt]

No I mean I nicknamed the Texture2D as sprite

In the shader graph

[fossil cedar]

but in the "old days" have the alpha mask be baked into an rgb texture as bright green or magenta, etc was common

so that's what i'm wondering but just guessing

[lucid geyser]

Check this out. so all tuts have showed something similar to this when creating a brand new mater shader. the only different is this whole "fragment" and "vertex" thing is't included. i'm thinking i might be using the wrong kind of shader as a starting point?

the tuts show somthing like a "sprite lit master" which i don't have

notice i'm plugging a v4 color into a base color 3 and taking the alphga and pluggin gthe alpha 1. that seemed to resolve the greeen parts of the sprite.. but i'm thinking the yellow issue has something to do with the weird fragment and vertex starting point

[fossil cedar]

yeah they changed master nodes to vertex and fragment, but the shader graph has a certain type and target when you create it and in the properties like "lit" or "unlit" "decal" etc

sorry for linking to an old article

i'm testing something quick

[lucid geyser]

ahhh wait... do i just need to plug in a color node into the base color 3?

i think this is working

[fossil cedar]

that's one way to do it anyway if that's all you need. still weird this yellow green color

@lucid geyser i was going to say try this on the texture import settings

change from sprite to default

but that's just a guess

[lucid geyser]

ok i got it to work...

[fossil cedar]

it maybe makes no difference, maybe sprite is just another way to tell the editor you want have the option to edit the texture pixel by pixel but doesn't change the texture format, so more to the point of what @clear veldt was saying.

[clear veldt]

wdymmmmmmmm

[fossil cedar]

i haven't done much 2D sorry for going on with some wild guesses

[fossil cedar]

when i select sprite as texture type it says i can install the sprite editor and edit sprites

anyway, sorry off topic

[clear veldt]

Oh, I haven’t had that

[clear veldt]

But I’m working on 3d

[fossil cedar]

@lucid geyser even though you worked around the issue the yellow green seems strange. i would try restarting unity or restarting your computer at some point, and try updating the GPU driver to see if it persists or goes away. not exactly the same issue just little dots but bright yellow and green. the person updated their GPU driver software and it went away https://forum.unity.com/threads/weird-green-and-yellow-pixels-appears-on-textures.386844/

[fossil cedar]

oh yeah true, well i haven't worked with sprites in the sense of unity's texture type either and they seem geared towards "(2D or UI)" according to the dropdown. so i guess that could be UI in 3d game. also is think particlesystem uses sprite type? but i mostly use vfx graph for particles and "default" texture type

[clear veldt]

Ok

[tranquil jackal]

thank you I will try to do that. I am using Memory profiler, and checking what it says.

[fossil cedar]

this sounds likes normal default behavior in scene view in edit mode. unity tries to optimize some things in the editor to not run unless you are interacting with them. there are a few things you can do, enter play mode, and/r make sure "always refresh" is enabled in scene mode setttings, it's in a drop down that looks like a sparkle for "effects." there are also some editor scripts to force the editor to repaint every frame but they aren't really officially supported so might have strange results

[fossil cedar]

i had a response to you about this already but i lost it i have to retype it lol. i was going to say, i'm not sure what your goal is, but it *looks *like you are (rail) "road building" and in that case i recommend using the Unity Spline tool package for this "road / world bending" on the CPU / in C# script land which will have fewer bugs and weirdness down the road (literally) than doing large movements of the game world "en masse" via on the GPU in shader land.

vertex position offset shaders like this should generally only be used for small movements like leaves blow in the wind, or MAYBE for some large movements that are only visual effects in the background like a thousand birds flying in the background that you never touch or electricity beams or sparks that fly right through you but you don't have any game logic attached to them. or maybe a weird world bending cinematic.. even then i would probably use the Mesh API and a script or compute shader instead of a vertex shader

assuming splines are the way here i think #🔀┃art-asset-workflow is the place to discuss the Unity Spline Tool further, i can see people have discussed it there.
https://blog.unity.com/engine-platform/building-better-paths-with-splines-in-2022-2
https://www.youtube.com/watch?v=CyN8Rw18_zw
https://www.youtube.com/watch?v=kNMu83sU948

[clear veldt]

Im not using splines, I have an endless runner so I just use a shader

My worlds already bends, some objects just float around

[fossil cedar]

@clear veldt ohh because it's endless runner, the world coming at you is 3D but the player only interacts what's essentially a 2D plane. so it doesn't matter gameplay wise if the world bends in front of you with a shader. sorry for misunderstanding.

@clear veldt in this case i noticed a few months ago Unity released an official endless runner game example. i looked it up and it says it includes a "curved world shader". perhaps we can use this shader or at least cross reference it to your shader graph to help resolve the issue. https://assetstore.unity.com/packages/templates/tutorials/endless-runner-sample-game-87901
this tutorial related to the example says "The files to examine to see how this works are WorldCurver.cs, and CurvedCode.cginc"
https://learn.unity.com/tutorial/mobile-development-techniques#5c7f8528edbc2a002053b4ab

[lucid geyser]

How can i make my shader sprite render behind other sprites?

[fossil cedar]

for 2D?

[lucid geyser]

yes, it doesn't behave the same as othe rsprites

[fossil cedar]

https://docs.unity3d.com/Manual/2DSorting.html

there's a number of ways. i think that doc lists all of them

[lucid geyser]

this doesn't help.

I want to speak with a person about it.

[fossil cedar]

i'm a person lol but i don't work for Unity. it's just a community forum

responding immediately with "this doesn't help" means you didn't even try. (or if you did find and try these methods earlier, i guess you aren't saying what happened or why it didn't seem to work?)

i recommend trying the first two methods in the doc. either put the sprite in a sorting layer and rearrange the layers so yours is in the back, or set the render queue for that renderer to a number lower than the other renderers

1. Sorting Layer and Order in Layer
2. Specify Render Queue

there's screenshots in there for how the sorting layers work even

it's linked from the 2D sorting article above https://docs.unity3d.com/Manual/class-TagManager.html#SortingLayers

[clear veldt]

Oh I think I’ll look at it

[lucid geyser]

Does anyone know how to have multiple objects use the same shader but behave indipendantly of one another?

[uncut inlet]

why do my textures look like this after sampling

[vocal narwhal]

you are presumably ignoring the alpha channel

[shadow locust]

Give them distinct materials that use the same shader

[vale spruce]

is there any performance cost of Shader.SetGlobalMatrix rather then setting them one by one ?

this tutorial telling me to use two kind of struct v2f, and for some reasons it works, but I don't understand how

[green peak]

structs are just objects

you can have multiple

[vale spruce]

InterpolatorsVertex MyVertexProgram (VertexData v) {
    InterpolatorsVertex i;
    UNITY_INITIALIZE_OUTPUT(Interpolators, i);
    UNITY_SETUP_INSTANCE_ID(v);
    UNITY_TRANSFER_INSTANCE_ID(v, i);

    i.pos = UnityObjectToClipPos(v.vertex);
    i.worldPos.xyz = mul(unity_ObjectToWorld, v.vertex);
// rest of code
}

[vale spruce]

but what I confused is that both use same name variable

both struct variable name is i

[green peak]

doesnt look like it

InterpolatorsVertex MyVertexProgram (VertexData v) {
    InterpolatorsVertex i; // <----- I see only one!!!!
    UNITY_INITIALIZE_OUTPUT(Interpolators, i);
    UNITY_SETUP_INSTANCE_ID(v);
    UNITY_TRANSFER_INSTANCE_ID(v, i);

    i.pos = UnityObjectToClipPos(v.vertex);
    i.worldPos.xyz = mul(unity_ObjectToWorld, v.vertex);
// rest of code
}

[vale spruce]

but then what is the Interpolators variable name?

[green peak]

based on the code you showed me, its not declared at all

 UNITY_INITIALIZE_OUTPUT(Interpolators, i);

what does that function do, i would guess your answers lies there

[vale spruce]

oh... now that I look at it seems like the turtorial actually split the v2f vert (appdata v){} into multiple function

[green peak]

hahaha, catlike coding is it?

He does that, gives you the answer before the question and ur left guessing unless you read along

I recommend you do a quick read of the full page, then come along a second time

[vale spruce]

I see... thanks, I guess I need to reread it again

[green peak]

Anytime soldier, we are all in this battle together!

[karmic hatch]

If you remove the inverse model and switch everything into world space (including the vertex node), that might fix it

[karmic hatch]

Shadertoy (online) has lots of artistic shaders like that, they're in GLSL but the main difference is just that HLSL calls its vectors float234 while GLSL calls them vec234

[green peak]

not a shader question but,
you already did?

[stable cove]

i'm sorry wrong channale

[clear veldt]

I’ll try

[frail violet]

Greetings. I'm using standard pipeline on Unity 2020.
I'm trying to make a shader for a material on a cube.

I want the top face to be a certain texture, and I want the side faces to be a different texture.

I want it to tile based on x and z scale (but only on the top face).

I want the textures to be emissive. I want to be able to control the color, and intensity (values greater than 0).

I found some examples online, but they were all based on world position, and couldn't figure out how to do it just based on the local normals and scale of the cube.

[shadow locust]

YOu can do this with:

• multiple materials (requires your mesh is broken into submeshes)
• UV mapping (you will need to UV map your mesh such that each face uses a separate UV coordinate section)

With the UV mapping approach you can either combine all your textures into one, or do a mapping in your code where you figure out which UV section the fragment is in and then normalize the UV to fully map it to one of the input textures

[frank sparrow]

can anyone give me some advice, i am trying to make a shader graph that is just turns a plane into a moving sine wave. ive experimented a load and i just cant seem to do it and cant figure it out. i have been told this is a relatively simple vertex shader but all i end up doing is changing the position of the entire plane or scaling the plane so I cant figure out how to individually change the vertexes to map to the sine wave. any pointers would be appreciated thank you

[fluid salmon]

hey everyone! I have a beach shoreline inspired by Cyanilux's animal crossing twitter thread.

I have the waves flowing like in the first image in time. I'm using the second image as a mask to make it seem like the waves start off separate and combine to form one. However, I'd like to make it so there is variance. I dont want all the waves to start off as the same size. Does anyone have ideas as to how i could do this?

[shadow locust]

make sure to include the vertex's position in the calculations somehow or else all vertices will be affected the same

basically somethinbg like this is what you want to express in the graph, as an example:

vertex.pos.y = vertex.pos.y + Sin(time + vertex.pos.x)```

depending on what effect you want you may include the z position in there somehow too, or not

[frank sparrow]

alright, is there a node to access vertex position? i presume it isnt vertex id/color

[shadow locust]

yep, the Position node

[frank sparrow]

im trying to make a form of water for my boat project so looking at using gerstner waves eventually

[frank sparrow]

is it one of these options or do i need to add extra nodes to the position node to access vertex positions?

[shadow locust]

I'm not familiar with gerstner waves but there seem to be many videos about that online

[shadow locust]

object is correct

or world

[frank sparrow]

yeah i understand the mathematical concepts just trying to understand shader graphs now in order to oimplement them

[shadow locust]

basically it depends where this node ultimately feeds into:

• if you're feeding it into the vertex stage it means the position of the vertex
• if you're feeding it into the fragment stage it means the position of the fragment

we want a vertex shader here

[frank sparrow]

im feeding it into the Postion bit on the Vertex

okay ill try test around more, ive been using the object position node before so ill see what i can do

[frank sparrow]

im still not sure how to fix this. this shader graph is creating a flat plane and moving it up and down in a non linear scale and scaling it larger and smaller. also i tried using sine time and the same thing occured. i tried using the idea you wrote out here in code

[shadow locust]

the image is a bit too blurry for me to read properly

even when I "open in browser"

[frank sparrow]

oh sorry

let me get a better screenshot

any better?

[shadow locust]

you're adding the whole position to the time node

try adding just x

[frank sparrow]

oh crap yeah

is that just a single node or something else?

[shadow locust]

you can use https://docs.unity3d.com/Packages/com.unity.shadergraph@16.0/manual/Split-Node.html

[frank sparrow]

does that mean i could select any two of those and they would be equal to x and y since the vector position is less than four inputs?

not any two actually, the first two and that would be X and Y

[regal stag]

The position is a Vector3, so there's also a Z component (aka B). RGBA=XYZW

[frank sparrow]

what would the W component be?

oh my god its so beautiful

[regal stag]

In this case, for the Split it should just be 0 as there is no w component for a Vector3.

[frank sparrow]

ah okay thats good

thank you soo much PraetorBlue and Cyan, i have done it, now for some minor adjustments

[regal stag]

Also to add to Praetor's suggestion, you might want the result of the Sine here to only offset vertices along the vertical/Y axis. Can put it into a Vector3 node before adding to the original position to do that.
Though something to keep in mind for later, when doing actual Gerstner wave calculations they usually have some horizontal/XZ offset too.

[frank sparrow]

i didnt know i could store them in vector3 variables, thank you very much. gonna be much easiser now i think

[grave vortex]

Just bumping my previous question

Clearly using Nsight to profile unity shaders is possible, as evidenced by this video, but I haven't been able to find any information on how to export shader symbols with a build

https://www.youtube.com/watch?v=l_LiE1vAFhM

[fluid salmon]

hey Cyan! In your old animal crossing shoreline shader (https://twitter.com/Cyanilux/status/1251899743644549122) you have the waves starting off separated and combining as they get closer to the shore. I assume you're using a mask like the one i've attached. How would you make it so that the distance between these separated waves varies? So that the waves dont all combine at the same time?
I'm thinking offsetting the Y coordinate of that texture based on some randomness?

[regal stag]

Hmm, iirc the mask I used was some worldspace mapped noise. Probably lerped to white over the UV.y so it leaves less gaps/holes closer to the shore.

[drowsy coral]

Hey everybody, this is with built-in rendering pipeline.

I'm trying to set fog colors from a dictionary.

Setting the fog color beforehand in the editor and not modifying it in code comes out fine:

But as soon as I set it in code, look what happens, no matter what setting I change.

The fog is pure white, and I can't even go change it in editor! (Granted, it's because I'm changing it once per frame in an update function, but still).

When I debug, too, the fog is being assigned properly...

But it comes out as white.

Also I start getting depth errors with my billboarded sprites, but that's a whole nother issue.

[celest citrus]

Hi! I need to create a shader using shader graph that will procedurally generate a space background with stars. The main problem is that I have to use sprites of stars for this.

I will be glad if somebody could explain how to do this or where to search for guides.
Thanks!

[polar ruin]

Hi, what would be the command to toggle the doubleSided property at runtime?
I'm not sure where to look. documentation points towards this but there is no mention of it :/

https://docs.unity3d.com/Packages/com.unity.render-pipelines.high-definition@16.0/api/UnityEngine.Rendering.HighDefinition.HDMaterial.html

[wary axle]

I want to write shader in unity, but it does not create an error in unity, but it gives an error through visual studio, how do I solve it (visual studio is defined in unity)

[clear veldt]

Didn’t work, and wdym by vertex node?

[karmic hatch]

Instead of sampling a sine wave you could try sampling gradient noise (or just two sine waves added together with different frequencies)

[karmic hatch]

The node it feeds into with the vertex position, normal, and whatever else

If the vertex position input is still set to object space in the node inspector it's going to break everything when object space isn't world space (i.e. when something rotates) but if everything is in world space it should work

[frank sparrow]

is there any reason I cannot connect these nodes

they have been connected in the past and now I cant reconnect them

[regal stag]

Chains of nodes sometimes get "locked" to a specific stage (vertex/fragment). It's usually recommended not to mix them

[frank sparrow]

ah alright thank you

is there a way I can take the height of the vertexes used in a shader graph and use them in a script?

[regal stag]

Not really no. I assume you want it for something like bouyancy if this is still for water. The usual way is to replicate the displacement calculations on the C# side.

[frank sparrow]

ah crap alright

thank you

i have decided i hate gerstner waves too

[ebon moss]

Hey, I have a wind effect shader graph and I'm trying to turn it into a wind vignette (wind effect around the edges of the screen), any ideas how I coulod do this

[kindred crystal]

hey there

having an issue with a custom character shader, on mac it doesn't cast shadows and ambient occlusion draws on top of the character. on windows it's fine. seems like something going on with the camera depth buffer? anyone ever seen a thing like that?

[kindred cradle]

Anyone know why my exposed shader graph properties aren't editable from the Unity UI?
In the image I have selected "Size" which shows it is "Exposed" from the Inspector but in the material everything is grayed out except for the Colors.

[kind juniper]

Where is the material asset located?

[kindred cradle]

Hopefully youre talking about the filesystem, if so its /Assets/Shaders/WaterShader

[kind juniper]

Hmm... It just looked like a default unity asset to me being uneditable, so I assumed it's in some kind of protected folder. But I guess that's not the case.

[kindred cradle]

I tested this on a Core RP converted to URP and a brand new URP project, saved in different areas so idk whats going on man! 😂

[regal stag]

Shader Graphs automatically create a Material asset under them (in 2022+), which is uneditable but uses the defaults from the graph. If you want a material with different values you need to create a separate Material asset.

[kindred cradle]

Wow youre so smart, thank you
And thanks for trying dlich!

[kind juniper]

Aha, so it was a "default asset". Didn't know that the shader graph creates a material for you now.

[split thorn]

anyone know if it is possible to get a shader range property min value and max value?

[kindred crystal]

is there an issue w/ using a custom shader uniform in a depth / shadowcaster pass?

set on the material e.g. material.SetFloat

as soon as i reference any of those uniforms, the depth pass for this material gets dropped. not in the framedebugger.

[vale spruce]

Anyone could think a better way to write this?

float xCoordsColor = i.customShadowCoords.x > 1 ? 0 :  i.customShadowCoords.x < 0 ? 0 : 1;
float yCoordsColor = i.customShadowCoords.y > 1 ? 0 :  i.customShadowCoords.y < 0 ? 0 : 1;
return xCoordsColor * yCoordsColor;

I just want it when i.customShadowCoords.x and i.customShadowCoords.y both value is between 0 to 1, it will give me 1 as value, else it will give me 0

[kind juniper]

Clamp..?

[vale spruce]

clamp? how do you use clamp for this?

[kind juniper]

Nvm. Misread the code.

[vocal narwhal]

Here's my lazy and shitty first thought:

float x01 = saturate(ceil(0.5-abs(x-0.5)));
float y01 = saturate(ceil(0.5-abs(y-0.5)));
float result = x01 * y01;

[vale spruce]

yeah... looks more expensive probably no 😅

[vocal narwhal]

I doubt it's more expensive, it's branchless 😄

[vale spruce]

hmm... maybe you are right

[vocal narwhal]

1 - saturate(abs(floor(x))) also an option for the equation.
You can have a play around with something like graphtoy and see if you can get it any better

[vale spruce]

I see... thanks

[vale spruce]

float result = 1 - saturate(abs(floor(i.customShadowCoords.x)) + abs(floor(i.customShadowCoords.y)))
ok found the answer

[amber saffron]

An other way ? 🙂

float2 t = uv;
t = step( abs(t-0.5), 0.5);
float result = min(t.x, t.y);

[vocal narwhal]

I was trying to avoid step. but there are lots of ways 😄

[amber saffron]

Any reason why ?

Similar without step :

float2 t = abs( uv - 0.5 );
float r = max(t.x, t.y);
r = 1.0-floor(r*2.0);

[vocal narwhal]

🤷 min(step(0, x), step(x, 1)) just chain a whole bunch of steps together, then you have the same solution has the first one but with a fancy function instead

[frank sparrow]

In the position node, what does the Object and World actually correlate to?

[vocal narwhal]

Object space and world space

[frank sparrow]

should have made myself clearer, what is object space?

[vocal narwhal]

https://unity.huh.how/programming/vectors/positions-and-directions the only part of this page on vectors I'm making covers just that. It's the same as local space

[frank sparrow]

ah alright ill give it a read thank you

okay i think i get it thanks

[civic oxide]

Does anyone know of a free dynamic culling system I can use that works with hdrp?

[amber saffron]

What type of culling ? Like occlusion culling ?

[civic oxide]

yup

[amber saffron]

This ? https://docs.unity3d.com/Manual/OcclusionCulling.html

[civic oxide]

I need it to be bakable during runtime and the only free solution I've found so far is onot compatible with hdrp

[amber saffron]

Okay. Well I don't know any runtime solutions, sorry

[frigid jay]

There is Entities.Graphics Burst Occlusion. It is part of ECS but it is open source if this matters.

Can give you insights maybe

[kindred crystal]

sorry to be a bother, this issue is driving me crazy. i set up a windows dev environment to see what the frame debugger looks like on that side. i boiled down the shadowcaster to something minimially reproducible (well, in our project anyways):

Pass {
    // -- options --
    Tags {
        "LightMode" = "ShadowCaster"
    }

    // -- program --
    CGPROGRAM
    #pragma vertex DrawVert
    #pragma fragment DrawFrag
    #pragma multi_compile_shadowcaster

    // -- includes --
    #include "UnityCG.cginc"

    // -- props --
    float1 _Distortion_Intensity;

    // -- types --
    struct FragIn {
        V2F_SHADOW_CASTER;
    };

    FragIn DrawVert(appdata_base v) {
        FragIn o;
        TRANSFER_SHADOW_CASTER_NORMALOFFSET(o);
        o.pos.y += _Distortion_Intensity;
        return o;
    }

    float4 DrawFrag(FragIn IN) : SV_Target {
        SHADOW_CASTER_FRAGMENT(IN);
    }
    ENDCG
}

on windows, the camera depth pass for this material appears in the frame debugger, the shader property is bound correctly

on mac, w/ the same exact code, this DrawMesh frame is missing from the depth pass unless i remove the reference to _Distortion_Intensity

[grave vortex]

Anyone know how to export shader symbols in a build?

(Just bumping my previous question again)

[frigid jay]

I don’t remember that it works for Nsight, but you can try #pragma enable_d3d11_debug_symbols

[grave vortex]

I did see that in the documentation, but it didn't seem to work.
The documentation also mentions it generates symbols for Vulkan, which is what I'm currently using, but maybe I should try a d3d11 build come to think of it

I was using Vulkan for Nsight GPU trace, as it doesn't support D3D11

[frigid jay]

If you are targeting Vulkan you can try to use newer DXC compiler: https://docs.google.com/document/u/0/d/1yHARKE5NwOGmWKZY2z3EPwSz5V_ZxTDT8RnRl521iyE/mobilebasic#

[grave vortex]

That's a good idea actually, I'll give both a d3d11 build, and a DXC Vulcan build a shot

[lone crow]

Why isn't the material with emission on isn't emitting any light onto surrouding objects?

[frigid jay]

Emission materials by default represent materials that not follow to light energy conseravtion principle. I.e. those materials can reflect more light energy then recevie from all surrounding light sources. Those materials are not light sources. For more realistic look with multiple light energy bounces (in other words light energy emitted from object is applied to surrounding environment) you need some form of Global Illumination (GI)

[lone crow]

So If I add a mild light source and raise the intensity value of the color on the material, that should make the material emit green light right? Or it woudl appear that way

small amount of light falling on the material which doesnt affect the cube much but the material will reflect more light than what is being received because of high emission intensity

is this how it works?

[shadow locust]

no you need to use global illumination

in other words

baked lighting

and it's generally not going to be realtime

[lone crow]

I have realtime global illumination on

[shadow locust]

I think to receive realtime lighting from this, you probably need light probes?

https://www.youtube.com/watch?v=2PBgCl-zIZQ

even in this video he uses a point light i think to get it to actually work in realtime

[lone crow]

Ahh thanks

Now its clear

[kindred crystal]

can confirm after inspecting the compiled output for d3d and metal and they both include a shadowcaster pass compiled for the same set of keywords. on mac, the shader inspector lists "Cast shadows" as no. and if i log the number of passes in the shader, on mac in reports 1 and on windows it reports 2:

Debug.Log($"[shader] {shader} subshader {shader.subshaderCount} pass {shader.GetPassCountInSubshader(0)}");

wish unity were open source

[kindred crystal]

pretty much at the end of anything i can imagine trying, running into the brick wall of whatever is happening behind the c# <> native shader bindings. no idea why unity is dropping the shadowcaster pass on mac.

[split oyster]

How do I render 1 milion of 2D circle sprites facing always the player?

[lunar valley]

indirect gpu instance 1 million quads and make them face the player inside the vertex shader

[split oyster]

I have a problem with indirect gpu instance... I was following a guy tutorial but I didn't understand a thing
for set vertex positions... have I to use a matrix?

struct appdata_t {
                float4 vertex   : POSITION;
                float4 color    : COLOR;
            };

            struct v2f {
                float4 vertex   : SV_POSITION;
                fixed4 color    : COLOR;
            }; 

            struct MeshProperties {
                float4x4 mat;
                float4 color;
            };

            StructuredBuffer<MeshProperties> _Properties;

            v2f vert(appdata_t i, uint instanceID: SV_InstanceID) {
                v2f o;

                float4 pos = mul(_Properties[instanceID].mat, i.vertex);
                o.vertex = UnityObjectToClipPos(pos);
                o.color = _Properties[instanceID].color;

                return o;
            }

I wanna use a float3 instead of matrix and float4 stuff... is it possible? I heard that float for vertex is necessary for not break the rendering... so is there a way to efficiently convert float3 to float4 in shaders?

[tacit hearth]

does anyone happen to know of a shader graph node that will let me colorize an inputted texture? i want to be able to colorize images white but multiplying by the color white obviously just uses the original color from the image

[kind juniper]

How exactly do you want to "colorize" it?

[tacit hearth]

by changing all the colors in the image to white, same thing you can do to textures in image editors

[kind juniper]

If you change all the colors to white, you'll get a completely white texture though...🤔

[tacit hearth]

or shades of

[kind juniper]

Do you have a specific example of it from an image editor?

[tacit hearth]

well colorizing in image editors does different shades according to the strength of the color, so i suppose its similar to trying to create a heightmap from a color texture

but ill screenshot 1 sec

original:

colorized white / grey:

[kind juniper]

I think that's just decreasing the saturation of the image.

[tacit hearth]

ah really? ill try a saturate node then, cheers

[kind juniper]

No, saturate in shaders is a different thing afaik. It just normalizes the values.

[tacit hearth]

ah

hmm, not sure which node would fit the bill....ill look about

[kind juniper]

I think you can use Colorspace conversion node to convert to HSV, then set the S(saturation) to 0 and convert back to rgb.

[tacit hearth]

this mayhaps?

[kind juniper]

Ah wait, there's actually a "Saturation" node as well

[tacit hearth]

aha!

that may be it

[kind juniper]

This just replaces the colors. I'm not sure that's what you want.

[kind juniper]

Yeah.

[tacit hearth]

thanks, i knew saturate existed....but not saturation XD

this opens up wild possibilities 😛

[vale spruce]

so I'm getting baked shadow by doing

indirectLight.diffuse = DecodeLightmap(UNITY_SAMPLE_TEX2D(unity_Lightmap, i.lightmapUV));

but what I actually want is toon like shadow, so... is there a away to bake like toon shadow (without any gradient)?

[amber saffron]

The lightmapper can't be customized (at least, not unity build in one) and uses some standard pathtracing to render GI and shadows realistically.
But to have a toony look to your shadows, you could apply some ramping to them, similarly as what is done in regular toon shading.

[vale spruce]

I see... thanks

[tight phoenix]

in shader, is there any way to get/lerp over an objects bounds? As in like, have a shader effect isolated to its worldspace min/max bounds?

Ideally without passing in vector3s through a material property block I can do that, but I was hoping the shader has access to the bounds directly?

oh object node has bounds 👀

hrm

how do I "get" the bounds in object space?

i guess i could do two transform nodes for min and max

i feel this a more basic math function could achieve this however

the output is pure white, but I am expecting a gradient from min to max

what am I misunderstanding?

are min and max bounds broken? or straight up just don't do anything? I am not understanding why the output is a solid color

yeah im starting to think these just don't output anything of any kind, nothing I do produces any values, useless and broken, actively misleading, poorly documented may as well be flat out lying by claiming to exist and offer this thing

the distance between min and max bounds is not 1 at every fragment so this obviously 100% is completely broken

how do I get bounds in shader? The object node is broken and returns nothing usable

like am I just stupid? how can this return 1 for every fragment, that makes no sense

[grizzled bolt]

If your cube's min bound is -0.5,-0.5,-0.5 and the max bound is 0.5,0.5,0.5 then I imagine their distance would always be exactly 1

I don't recall if it was local space or world space but the distance would always be 1 nonetheless

[tight phoenix]

what about when im subtracting the fragment's object space position?

why would it be always 1?

I dont understand what I am doing wrong, I just want to have an effect isolated to the bounds of the mesh, no matter what bounds the mesh is

distance = destination - origin, but when I use min max bounds, I am not getting any distance

finally getting something, I still dont know why min max bounds isnt working or producing the result I expect

[grizzled bolt]

Bounds are two world space coordinates, so taking their distance doesn't produce a variable that would change per fragment

Like, they are points, not coordinate spaces

I may have something about this somewhere

[tight phoenix]

thats immensely confusing, how can the world space coordinates of min and max not produce distance? like, its positions in world space, ergo distance between those positions would differ per fragment???

I'm starting to think that I am the stupid one, that it does exactly what it does but im too dumb to understand how to use it

[grizzled bolt]

The distance of example bounds points -0.5,-0.5,-0.5 and 0.5,0.5,0.5 remains the same regardless of when or where you measure that distance

[tight phoenix]

so why is distance still identical even when I subtract object space positions from it?

that should introduce spacial coordinates if it doesnt already have that, right?

I mean obviously I am wrong because its not working but I dont understand why thats wrong

it also behaves completely different when done in world space coordinates instead of object space, which also makes no sense to me

world and object space coordinates are synonymous as far as I knew, the same calculation done in all world should return the same as done in all object space

I dont know why this is agitating me so severely and my agitation is only making it harder to comprehen

[grizzled bolt]

Hey I understood this before and now I don't!

[tight phoenix]

yeah

trying to show myself empathy and not get mad at myself for not understanding

[grizzled bolt]

I think it should be something like this, for a vertical 0-1 gradient
We just want the world position "stretched" between the lowest bounds point and the highest bounds point using remap
But something's wrong here because the gradient changes around even when rotating a sphere

It looks mostly right but when rotating a sphere the bounds positions shouldn't change at all, but it seems like they do

[tight phoenix]

here is where I am at, about to test a sphere like you suggested, I only did cube so far

the output of mine

Ah yeah Bounds is in world space with no rotation

meaning bounds grows as it rotates, its size changes 🤔

[tight phoenix]

oh yeah but it also changes on a sphere in mine as well

the bounds of a sphere do not change by rotating it (in theory) so that's pretty wtf

[grizzled bolt]

Totally!

[tight phoenix]

this is my guess at whats happening

that its treating bounds more like object space box collider even though it says its a world space min max

[grizzled bolt]

That explains the value change, and it's not incorrect that the coordinates themselves are in world space
But I didn't know that bounding boxes could rotate at all

[tight phoenix]

Yeah me either shaders can get the rotation, maybe we can do some kind of matrix multiplication to remove rotation before the calculation takes place?

[grizzled bolt]

I guess the reason is it refers Renderer.localBounds instead of Renderer.bounds

[tight phoenix]

oh and we DONT have access to the rotation either 😬

[grizzled bolt]

Oof
I'd find some lazier workaround
Or live with localbounds
It technically works anyway

[tight phoenix]

hm I am doing mine in manhattan distance which isnt technically working on my end

but I am not sure what kind of matrix multiplication I have to do to get all these values in the same space irregardless of any rotation or scale

i know matrix mults must be done, but not sure what matrix, and what elements, and when in the calculation

this looks close

[grizzled bolt]

@tight phoenix I think you can use this if you want a -1 to 1 space within your bounds on all axes, used in place of world/object position

So any noise nodes and such would be using these coordinates, not within placed among these nodes

[tight phoenix]

you're only using the x value of the min and max bounds?

I am getting broken texture and pure white output from your logic there

[grizzled bolt]

Oops that's right
needs more work

[amber saffron]

Using world position and bounds, you can do a simple remap yourself :
remapedvalue = (pos - boundMin)/(boundMax - boundMin)

[tight phoenix]

im also realizing that I cant use the position node at all because the position node has no concept of bounds, its just a value where 000 is the mesh dead center

[tight phoenix]

I will try this now

[tight phoenix]

am I doing this right? the output is broken, in scene there is sorta a gradient but its not in manhattan distance and I no longer know 'where/when' to convert the distance to that

I am trying to get box bounds of any mesh, regardless of scale, rotation, position
in manhattan distance values so that there is no visible 'sphere' to the distance

where 0 is one corner and 1 is the other

[amber saffron]

The preview is broken, but what about the scene view ?

[tight phoenix]

in scene there is sorta a gradient but its not in manhattan distance and I no longer know 'where/when' to convert the distance to that

[amber saffron]

BTW, if I'm not wrong, the manhattan distance should be the sum of XYZ of the divide node.

[tight phoenix]

sum as in x+y+z?

[amber saffron]

Yes

[tight phoenix]

this is sorta-kinda working but rotation is still affecting the output

I think the problem is that its in world space coordinates, and I need local object maybe?

[amber saffron]

Here's what I got :
The last divide it to normalize the gradient over the bounds (else the diagonal length is > 1)

[tight phoenix]

ill try matrix convert after the divide because its still a vec3 there

[amber saffron]

Yes, but shadergraph doesn't provide local bounds out of the box, you will have to pass them through script

[tight phoenix]

this broke it, so this was not the answer

[tight phoenix]

can't local bounds be derived from the world bounds?

[amber saffron]

No, it is only the other way around

The local bounds is a bouding box aligned with the object rotation, and the world bounds is a box containing all the local bounds corners, aligned with the world axes

[tight phoenix]

if world bounds min max change when the object rotates, can't you just do some kind of matrix multiply to remove that rotation before doing the calculation?

since world bounds and object bounds are the same if the object isnt rotated

[amber saffron]

But it will still not be the local bounds

[tight phoenix]

Why wouldnt it? This is wrong I assume?

[amber saffron]

Sorry, I answered before this sentence. Indeed, when the object isn't rotated they should match exactly

And it should work :

[tight phoenix]

does that account for rotation somehow?

I don't know how to get rotation in shader, net says complex things about matrix multiplication

[amber saffron]

Yes, transform node accounts for rotation

[tight phoenix]

okay, trying that now

[amber saffron]

Now, it really only works arount the 0 rotation, you get axes inversion and other weird stuff happening when you rotate :/

[tight phoenix]

output is still kinda wack and changes based on rotation

yeah I think the problem is that the world bounds are physically getting larger as it rotates

so it cant just be multiplied out since its bounds got bigger and not just changed spacial position

[amber saffron]

Because the world bounds change with rotation, and the world -> object transformation makes their values jump all over the place when you rotate

[tight phoenix]

Yeah that

so I guess the sollution is.. don't rotate the mesh which is not ideal

[amber saffron]

Just grab the local bounds from script and send them to the shader

[tight phoenix]

yeah I guess Ill have to do that

I am making use of a lot of material property blocks in my shaders and Im getting concerned that when all of this is in a scene together that performance might tank

but maybe it wont and my worries are unfounded

https://cdn.discordapp.com/emojis/556735450622984202.webp?size=96&quality=lossless

now I just have to figure out which iof these 5 attempts is made to make use of the correct bounds

and then make use of those bounds correctly

well that works for scale and rotate, but not distance

[stable nexus]

Is there a way to get screen depth in standard render pipeline?

[tight phoenix]

I have min max bounds positions in world or object space now, but I can't figure out how to get back to where I was to get the manhattan distance 😬

immensely frustrated

frustrated to the point that I know I cannot solve this without help

and in knowing that, its basically prophesizing and becomes true because I wont solve it without help, I cant, Ill stop myself

or something

anyways ill just take lunch

[amber saffron]

docs.unity3d.com/Manual/SL-CameraDepthTexture.html

[amber saffron]

If you have the min and max bounds in object space, it should be hard to get the gradient you want.
Can you share how you got this ?

[tight phoenix]

    private void PassBounds()
    {
        if (meshFilter != null && meshFilter.sharedMesh != null)
        {
            Bounds bounds = meshFilter.sharedMesh.bounds;

            // Object to World space.
            Vector3 cornerMin = transform.TransformPoint(bounds.center + Vector3.Scale(bounds.extents, new Vector3(-1, -1, -1)));
            Vector3 cornerMax = transform.TransformPoint(bounds.center + Vector3.Scale(bounds.extents, new Vector3(1, 1, 1)));

            // Object space;
            //Vector3 corner1 = bounds.center + Vector3.Scale(bounds.extents, new Vector3(-1, -1, -1));
            //Vector3 corner2 = bounds.center + Vector3.Scale(bounds.extents, new Vector3(1, 1, 1));

            matBlock.SetVector("_BoundsMin", cornerMin);
            matBlock.SetVector("_BoundsMax", cornerMax);
            Renderer renderer = GetComponent<Renderer>();
            renderer.SetPropertyBlock(matBlock);
        }
    }```

[amber saffron]

So, these are the object bounds world coordinates.
Note that instead of doing the maths, you could just use bouds.min and bounds.max

You'd better go to pass the bounds "raw" value and do the math in shader in object space

[devout coral]

anyone wanted outline for 2d sprite? How its made

[grizzled bolt]

Oh so that's why they change when a sphere is rotated
That piece of info I was missing

[tight phoenix]

good point

[tight phoenix]

oh missed this originally, also good to know

    {
        if (meshFilter != null && meshFilter.sharedMesh != null)
        {
            Bounds bounds = meshFilter.sharedMesh.bounds;
            matBlock.SetVector("_BoundsMin", bounds.min);
            matBlock.SetVector("_BoundsMax", bounds.max);
            Renderer renderer = GetComponent<Renderer>();
            renderer.SetPropertyBlock(matBlock);
        }
    }```
In this version, https://docs.unity3d.com/ScriptReference/Bounds.html per this, what 'space' are min and max in?

the math performed in transform.TransformPoint(bounds.center + Vector3.Scale(bounds.extents, new Vector3(-1, -1, -1)));
is a black box to me

I dont know what those values are or how to replicate them in-shader, I could guess based on their names but I wouldn't be able to debug that if it didnt just work

[amber saffron]

There is no notion of space in the bounds object itself, it just represents a box

[tight phoenix]

how can a box exist without space

even the term box implies space, my brain is short circuiting trying to comprehend of a box that has no space

[amber saffron]

Vector.Scale .... is just vector3 * Vector3
so this function transform a manually calculated bounds.min from local to world space

[tight phoenix]

its not a box, its just two coordinate positions that a box could be extrapolated from?

[amber saffron]

Well, yes. Kind of 😅
You could represent a box by the min and max, or the center and extend. But it can be summed up to two 3D vectors, indeed.

And by "no space", I meant it is not bound to any local object or world space in unity, this depends on the context where it is called from.

Renderer.bounds -> world space
Renderer.localBounds -> local space
Mesh.bounds -> relative to the mesh coordinates space

But all of them a just a bounds data struct

[tight phoenix]

spacial transformation and spacial coordinates and the lack of any coordinates at all are difficult 🫠

[tight phoenix]

I now have this, which allegedly is the bounds, in object space, fixed to the size of the bounds no matter its position, rotation, or scale

does that look right to you?

[amber saffron]

I think so.
Should give you local positions in the range -1;1

[tight phoenix]

seems correct yes pictured: stepping the output

okay now to get the manhattan distance from it somehow

I know to get manhattan distance I need corner to corner, and if the output is -1 to 1, I just need to add/subtract then manhattan those I think

[amber saffron]

From the center ? abs(position) and sum x+y+z

[tight phoenix]

why is this wrong?

why would the distance be always exactly the same no matter what fragment I evalulate if I am evaluating from positions that have been offset?

the problem is the offset positions were all offset by exactly 1 and -1.. so the difference didnt change?

But then how do you offset not all ofit by all of it?

[amber saffron]

What is happening in this custom function ?

[tight phoenix]

manhattan distance

[tight phoenix]

Out = abs(v1.x - v2.x) + abs(v1.y - v2.y) + abs(v1.z - v2.z);

[amber saffron]

Oh, wow, that's a lot of math for something simple 😅

[tight phoenix]

thats the result I got when I googled the formula for manhattan distance

how am I supposed to know its wrong :/

no where on the page said 'hey guess what this is actually wrong'

[amber saffron]

It is technically correct, but could be written in a more optimized manner

[tight phoenix]

Sorry Ill try to stop being snippy

optimized or not its still outputting a single value no matter what the fragment is

and I dont know why

unless the reason is because I am adding / subtracting to every value, this no change is created

[amber saffron]

Hum, you node setup would just output abs(range)*3 as manhattan distance here

[tight phoenix]

but then how am I supposed to add only to some fragments but not others

[tight phoenix]

normal distance is also a single solid value with no distance

[amber saffron]

What manhattan distance are you trying to get here ?

[tight phoenix]

min to max bounds extents

I wish I knew what I was doing, I wish this was easier :/

I wish I didnt have to rely on others to explain every single basic vector math 101

[amber saffron]

You are trying to get the distance from A+X to A-X => 2*X

[tight phoenix]

am I?

I have no idea wthat that means

is that the mathimatical formula that represents min to max bounds distance?

What is X in this case? Min, Max, and ?

I guess x is the fragment

[amber saffron]

I'm just trying to explain what is happening in the lasts nodes setup you've shared, and why it is a constant value

[amber saffron]

This doesn't make sense 🤔
min and max are value, and bounds extents is max - min

Again, what is the distance that you are trying to calculate.
From where to where ?

[tight phoenix]

you said the output is in a range between -1 and 1, ergo to get distance between min and max, I need to center my coordinates on -1 and +1 then get the distance between the two, so to move those positions over I added/subtracted Range from both, and then tried to get distance

[tight phoenix]

I already said, from min to max bounds extents

[tight phoenix]

said here when you asked me previously

vector +111 to vector-111, or whatever values are the min and max bounds

[amber saffron]

And I said that this has no sense. You can calculate the distance from the min to the max of the bounds (which is a constant), or the extends, but "min to max bounds extends" means nothing

[tight phoenix]

how can the distance between the furest point and closest point be constant

its a box

boxes are not constant width

This is so incredibly confusing, I have absolutely no idea why or how a box doesnt just have dimensions, a box is a 3d shape

I dont have comprehension

I dont have understanding

a box physically cannot NOT have dimensions, box by definition is a 3d shape, just like square is a 2d shape

Im sorry this must be frustrating that I don't understand

[amber saffron]

the box, like the bounds ? The bounds is a box, and it size is constant

[tight phoenix]

I have no comprehension on what you just said

you may as well be speaking darmok and jilad at tenagra, I understand each individual word but the do not add up to understanding to me

the box.. is bounds? bounds is a box? size? is constant?

How can size be constant

[amber saffron]

Bounds, also know as "bounding box"

[tight phoenix]

a box is a 3d volume, size and constantness dont even exist in 3d space in that sense

[amber saffron]

Bounds : a 3D box, aligned to some coordinates system, that encloses a mesh vertices

[tight phoenix]

Right, yes

and distance is also a trivial formula

[amber saffron]

Yes

[tight phoenix]

so why is this basically impossible

[amber saffron]

It is not

[tight phoenix]

then why havent we solved it already

[grizzled bolt]

If you measure the box from one corner to another corner, you'll always get the same result for that box
That we talked about originally

[tight phoenix]

if its not hard, why has it been hours

Sorry, I should stop whining, you're trying to help me

[amber saffron]

But in you node setup you're not calculating any valid distance, and you were unable to explain what distance you are trying to calculate

[tight phoenix]

unable to explain what distance you are trying to calculate
bounds is a box
boxes have dimensions
I want the distance from the minimum positino of the box bounds to the maximum

I dont understand why you keep asking me to explain what I want or saying I cant explain it

I want an output of values that go from 0 to 1 across the box from the min to the max of the bounds, in object space, also known as the distance

[amber saffron]

Ok, so you want the distance from the minimum bound value, to the maximum bound value ?
This is the extend X 2, and is constant

[tight phoenix]

that doesnt make any sense, this is where you're losing me

[amber saffron]

Ah THIS is something else.
You want the distance from the current pixel position to the minimum of the bounds

[tight phoenix]

I think the problem is you are telling me the exact literal answer to my question, and NOT telling me what I ACTAULLY want to know

but I cant ask you what I actually want to know because I'm too stupid to even know myself and if I knew i wouldnt be here

im sorry im inexperienced

[amber saffron]

I'll ellaborate on this one.
You already have the normalized position. In the range [-1;1]
You just have to transform it to the range [0;1].
A simple remap node or value = value * 0.5 + 0.5 will do the trick.

Then, from this, calculate the manhatan position (values are already all positive, so it is just x+y+z)

It will give a result that ranges from 0 to 3, so finish by dividing by 3 to normalize the value

[tight phoenix]

You were giving me the exact symantic answer my question, but my question itself was wrong, the words I was using to explain my queston were not the right words, but in my mind they were the right words because I had no possible way to know that I was not asking the question I was asking

[tight phoenix]

reading this now

[amber saffron]

I have an additional question : why the manhattan distance specifically ?

[tight phoenix]

because I want the values to act as a mask to an effect to resolve in all directions linearly rather than across the true distance where point to point is the furthest real distance

which, is also wrong, and dont listen to what I just said

because I do not have the language to explain what I want in words

and the words I just spoke do not represent what I want

I will draw it instead

because drawings are a language we both speak

that is manhattan distance, correct?

[amber saffron]

This can be done with min

[amber saffron]

(nope)

[tight phoenix]

https://media.discordapp.net/attachments/940458330458062879/976919694902706186/burningMaeSmall.gif

ffs

then is it Chebyshev distance

[amber saffron]

But you can do it with min nodes.
result = min(x, min( y, z ) );

[tight phoenix]

okay so the whole time I wanted Chebyshev and not Manhattan

every time I used the word manhattan, I meant Chebyshev

[amber saffron]

You know more mathemical term than me 😅

[tight phoenix]

yeah I dont evne know how to say that kind of distance without using its 'official' name

[amber saffron]

Note that your gradient in the picture goes from white to black.
If it was the other way around, you'd want do use max instead

[tight phoenix]

distance where every direction is the same distance? distance where.. diagonals are not normalized?

like I cant even describe it, its just 'chebyshev distance'

[tight phoenix]

Anyways this is a wild tangent now, I will try to impliment again based on this

[amber saffron]

Acoording to wikipedia : distance between two points given by the greatest single dimension difference

If I would to put in my wording

[tight phoenix]

this part - back on my earlier confusion, why is distance constant?
its because the min and max are positions, and the distance between them is a single value?
Every time I said min and max extents I meant its 3D world position to its opposite 3D world position, is the problem that the words I used to ask for what I want, the meaning wasnt interpreted the way I intended?

except what I just said above is ALSO wrong isnt it?

I didnt mean its 3D world position to opposite, I meant the fragment's position on the mesh... <something something something related to the two corners>

[tight phoenix]

which is what this is

I feel like the problem was ultimately not that I didn't understand, the problem is that I couldn't explain what I wanted correctly

the words I kept using were describing something different from what I meant to be asking for

unless thats wrong too

[amber saffron]

There is no "min and max extents".
For a bounds, you have two pair of intricated values : min / max and center/extends
The bounds for an object is constant (that why you only pass the bounds values to the shader a single time), so it will always result to the same value for every pixel

[tight phoenix]

because at this point im pretty much settled that im always wrong and know nothing

[tight phoenix]

...How??? How can there not be min and max extends? That already makes no sense right away

here is why my brain doesnt understand

a box has physical dimensions

I can touch two places on the box

ergo there can be a min and max

[amber saffron]

It's just terms and definitions.
Extents is the half diagonal of the bounds -> a dimension
min and max are positions

[tight phoenix]

the min being the lowest possible spot on the box, and the max being the highest possible spot

for a line its easy to visualize because lines have a start and an end, so for a 3D volume of a box its like rotating a diamond

[amber saffron]

"lowest" and "highest" in all 3 dimensions, but yes

[tight phoenix]

in this picture, is this picture depicting the min, the max, and the center?

[amber saffron]

Figure them as the corner with the lowest values in all dimensions, and the corner with the highest values in all dimensions

[tight phoenix]

which is 'extents' ?

[tight phoenix]

Yes exactly thats 100% what I understood this entire time

[amber saffron]

Half of the diagonal that goes from min to max

[tight phoenix]

What I am not understanding is why did it take this must time and effort to get to this point

like, It feels like we agree on everything, but then I go 'then x = x?' and then you go 'no, x = apple'

and I am just not grasping

thank you for your patience with putting up with me as well

min and max, lets pretend they're world coordinates, are (000) and (111)

[amber saffron]

Because there is so many weird stuff that is possible to do with shaders that if you don't state clearly the intend, it can easilly be misinterpretted

[tight phoenix]

which is what i understood, what you just said, lowest and highest possible

[tight phoenix]

yeah 🔥 😿

I should stop trying to 'understand' it for now and just get on with it

[tight phoenix]

I havent even tested if this works yet for example, by works i mean 'does what I want'

[amber saffron]

and center is (0.5, 0.5, 0.5) and extends is also (0.5, 0.5, 0.5)

[amber saffron]

I'm trying to explain why all that confusion.
Taking you example, min=(0,0,0) and max = (1,1,1).
Quoting you previously :

I want the distance from the minimum positino of the box bounds to the maximum

Well, to my reading, that's distance between min and max, so a constant square root of 3 🤷

The importand point was this :

I want an output of values that go from 0 to 1 across the box from the min to the max of the bounds, in object space, also known as the distance

Because the thing that is displayed across the box is the curent pixel, which is the element that has a variable position value.

[tight phoenix]

it seems to be working

[amber saffron]

So, are we all good now ? 🙂

[tight phoenix]

🫠 almost

I cant help but notice that after all that work, the outcome looks suspiciously like object position

is object space position ALREADY values 0 to 1 of the bounding box? did I just waste hours on something I already had?

[amber saffron]

For a single unit cube, yes, it is the same.
But for everything else more complex, it won't

[tight phoenix]

whew okay time justified

like this cylinder for example

huh interesting that its bounds corner is here

I guess it makes sense the bounds would be on the mesh and not on the box bounds?

or am I dumb and this entire time I wanted object position

[amber saffron]

Not exactly here, the bounds corner will be outside of the cylinder

[tight phoenix]

ooh yeah you're right

yeah this is exactly what I wanted

[amber saffron]

It is basically a remapped object position

[tight phoenix]

remapping its range to be 0 to 1 in object space instead of whatever length of real units it takes up

bounds-normalized, so to speak?

[amber saffron]

Yes 🙂

Saying that now, it could have been done with a single remap node 😅

Well, if remap was working for Vector3

Anyway, you got it working, now up to you to do something with this gradient

[tight phoenix]

yess

[tight phoenix]

remapping of what? 👀 💦

[amber saffron]

remaping the object position

[tight phoenix]

how would I remap its object position to 'bounds space'?

since 'bounds space' doesnt real

is it this bounds size value

[amber saffron]

sorry, by "object position", I mean "position node, object space"

[tight phoenix]

right, i understood that, the part I was asking is how can I remap this in a single step?

to get the same result?

since it sounds like if I can remap it in a single step that will be much more performant than passing bounds in through material property blocks

and then doing a shitload of math

[amber saffron]

You still need the bounds data, but it can be done with fewer nodes

[tight phoenix]

oh okay

the property block is the slowest bit so in that case nothing really changes

[amber saffron]

result = (position - min) / (max - min)

And that was my initial answer when I joined the conversation 😅

[tight phoenix]

and the only reason it didnt work was because we were using the min max from the object node, instead of its real min max from its real bounds

[amber saffron]

Yep

[tight phoenix]

Okay, the fact that I was able to understand why that was the case is reassuring

[tight phoenix]

yep result is the same

okay going with the optimized vesion, ill keep the old version as a backup just in case

[amber saffron]

Why the multiply by 0.5 after it ?

[tight phoenix]

Thanks for all the help, I know it wasnt easy to get through to me

[tight phoenix]

without it, the range is from corner to mid point instead of corner to opposite corner

Is this not the expected output?

[amber saffron]

No, it should just range from 0 to 1

[tight phoenix]

maybe because object space starts centered?

granted I had to do the *0.5 even before using object space

[amber saffron]

Nah, that's the whole point of remapping with the bounds min and max 😄

Can you post a readable screenshot of the rest of the connected nodes ?

[tight phoenix]

Yup 1 sec

[tight phoenix]

both required the 0.5 mult going into Split

original pre object space method output without 0.5 mult

[amber saffron]

Hum, nodes seems correct, so the only source of error I can think of are the bounds min and max values.
Did you change the code ?

[tight phoenix]

I dont think I did

that code runs OnValidate and Start so its up-to-date

[amber saffron]

Oo.
This is somehow unexpected.
For a defautl cube, it does display -0.5,-0.5,- 0.5 and 0.5,0.5,0.5 right ?

[tight phoenix]

oh uhh yes

erm

mnaybe?

in the picture, that is a default cube

[tight phoenix]

https://cdn.discordapp.com/attachments/497874081329184799/1133829951716605972/image.png
this picture those are all unity stock primitives

[amber saffron]

What did you set as default values for those variables ?

[tight phoenix]

oh before passing in? 000 and 111 for min/max

or do you mean in the script?

in the script I didnt set any default values at all

[amber saffron]

IDK, both, I'm confused, if the script is working properly it shouldn't display like this XD

[tight phoenix]

the script says this is bounds min and max

[amber saffron]

Hum, max is clearly wrong here

should be 0.5,0.5,0.5

this is the issue

[tight phoenix]

the issue is actually that I wrote BOUND without the s

[amber saffron]

HAha

[tight phoenix]

as usual, pebcak

[amber saffron]

Happens

[tight phoenix]

yep works now without mult

[amber saffron]

There should be an error logged in the console when the property does not exist

Nice, finally all fixed and working properly, now go do what you want with this gradient 😄

[lone crow]

Wanted to know how this works

We are multiplying position of a vertex by its normal?

[tight phoenix]

trying to make use of the output to create a directional mask that I can pass from point to point but so far I cant even seem to debug one
Im so @#$@^ing burnt out that I can't even do the most basic thing but im still on the clock at work even though ive done basically nothing deserving of being paid all day :/

iunno why im even posting this

i am miserable and cant stop feeling imposter syndrome

spoke too soon, its actually working exactly as desired right now

[final rampart]

wow

[tight phoenix]

I'm just so burnt out I can't even feel joy that I got it 🪹

[final rampart]

How can I make one of this that has the unity those squares

[tight phoenix]

if thats asking me how to make the above mask, I am not sure what you mean by that

[final rampart]

from the publisher

[tight phoenix]

hm the janked up edge appears to be occuring due to floating point problems, bounds to shader says its a weird shape

ill work on this more later because im too burnt out to know why anything is anything

[final rampart]

I mean that because of how square it looks

[tame topaz]

Your question doesn't make sense?

[uneven schooner]

So I've seen these before, know what they are for, have used them before, but was wondering.. does anyone know what these would be called?

can give context if needed

[grizzled bolt]

That could be helpful

[uneven schooner]

it's basically three images in one, red is one image, green is another, and blue is another.

you would get each different image by getting the different color values

[regal stag]

Would be called something along the lines of channel-packed masks I guess

[uneven schooner]

ohh

yeah that makes sense

https://youtu.be/jlKNOirh66E?t=267
Found a video containing a better explanation if needed (timestamp)

[grizzled bolt]

HDRP utilizes this notably
https://docs.unity3d.com/Packages/com.unity.render-pipelines.high-definition@14.0/manual/Mask-Map-and-Detail-Map.html
Functionally it's no different from separate textures, but cheaper

[uneven schooner]

I know what it's used for but was curious as to what it's actually called

[uneven schooner]

anyone know what this means?

only seems to happen if I input a Vector4 in UV.

but it still happens if I switch UV from Vector2 to Vector4

and this is the subgraph

and finally, this is line 58

just noticed these errors

[low lichen]

If you replace the Texture.Sample with Texture.SampleLevel and specify a mip level like 0, then the first error should not occur, which means the compiler won't unsuccessfully try to unroll the loops.

https://stackoverflow.com/questions/56581141/direct3d11-gradient-instruction-used-in-a-loop-with-varying-iteration-forcing

[steel drift]

Hi guys, can you help me with this URP shader? I have a URP Lit material, and my goal is to change the tile value of the '_DetailAlbedoMap' in runtime. I've implemented the following code to achieve this:
internal void SetTexture(Texture texture, int tileSize = 10) {
baseMaterial.SetTexture(detailAlbedoMap, texture);
var tilingScale = new Vector2(tileSize, tileSize);
baseMaterial.SetTextureScale(detailAlbedoMap, tilingScale);
skinnedMeshRenderer.materials[0] = baseMaterial;
Debug.Log(baseMaterial.name);
}

The problem I'm facing is that although the value gets updated, the changes are not being reflected in the game's renderer (mesh) until I make some small changes in the editor. Is this a bug, or am I doing something wrong?

[vale spruce]

how to override a function in hlsl?

like let's say I want to override BRDF3_Unity_PBS, only in one specific .cginc file is possible to do so?

[rare wren]

I am working on custom lighting/shadows in URP unlit shader graph. I have it working with the directional lights and point lights (I think point lights don't have this issue since the shadows are cast in 6 directions).

Spotlights give horrible shadow artifacts when the shadow is at the edge of the light's bounding box. Does someone know the reason for this and how to solve it maybe?

void AdditionalShadows_half (half3 WorldPosition, out half ShadowAtten){
    ShadowAtten = 1;
    #ifdef SHADERGRAPH_PREVIEW
        ShadowAtten = 1;
    #else
 
    uint pixelLightCount = GetAdditionalLightsCount();
    uint meshRenderingLayers = GetMeshRenderingLayer();
 
    // For Foward+ the LIGHT_LOOP_BEGIN macro will use inputData.normalizedScreenSpaceUV, inputData.positionWS, so create that:
    InputData inputData = (InputData)0;
    half4 screenPos = ComputeScreenPos(TransformWorldToHClip(WorldPosition));
    inputData.normalizedScreenSpaceUV = screenPos.xy / screenPos.w;
    inputData.positionWS = WorldPosition;
 
    LIGHT_LOOP_BEGIN(pixelLightCount)
        Light light = GetAdditionalLight(lightIndex, WorldPosition, half4(1,1,1,1));
    #ifdef _LIGHT_LAYERS
    if (IsMatchingLightLayer(light.layerMask, meshRenderingLayers))
    #endif
        {
            ShadowAtten *= light.shadowAttenuation;
        }
    LIGHT_LOOP_END
    #endif
}

[scarlet pulsar]

I have problem with setting Color to my shader. It doesn't changing.

var materialProperty = new MaterialPropertyBlock();
materialProperty.SetColor("_SelectColor",new Color(1,1,1,0.95f));
_lastMeshRenderer.SetPropertyBlock(materialProperty);

Changing for example SunlightLevel work fine.

[vale spruce]

why my cutout shadow caster? anyone got an idea?

[vale spruce]

how to make real time cutout shadow?

[meager pelican]

shadows are cast in a special pass that is rendered from the perspective of each light into a shadow map. Specifics might vary a bit by pipeline.

So you'd have to have the SHADOWCASTER pass render your object not as it's default, but as its cutout. So you can't use a default shadow caster pass, you have to include one in your shader. If using SG, you'd have to do that by hand, editing the resulting code-generated file, as I don't know of a way to do it in the tool (I think). That's why you're getting a quad, because that pass isn't doing the cutout calcs.

Custom geometry mods/discards are always a problem with shadows.

[vale spruce]

I do have them in that shader code, here the frag

float4 MyShadowFragmentProgram (Interpolators i) : SV_TARGET{
    UNITY_SETUP_INSTANCE_ID(i);
    #if defined(LOD_FADE_CROSSFADE)
        UnityApplyDitherCrossFade(i.vpos);
    #endif

    float alpha = GetAlpha(i);

    #if defined(_RENDERING_CUTOUT)
        clip(alpha - _AlphaCutoff);
    #endif

    #if SHADOWS_SEMITRANSPARENT
        float dither = tex3D(_DitherMaskLOD, float3(i.vpos.xy * 0.25, alpha * 0.9375)).a;
        clip(dither - 0.01);
    #endif

    #if defined(SHADOWS_CUBE)
        float depth = length(i.lightVec) + unity_LightShadowBias.x;
        depth *= _LightPositionRange.w;
        return UnityEncodeCubeShadowDepth(depth);
    #else
        return 0;
    #endif
}

The important part is just the clip(alpha - _AlphaCutoff); right? for this? but it didn't work for me on real time, but it does work on baked

[meager pelican]

Hmmm...IDK. Verify the pass name. What pass is that code in?

[vale spruce]

Pass
{
    Tags
    {
        "LightMode" = "ShadowCaster"
    }

that?

[meager pelican]

yes the clip

yes that

[vale spruce]

Unity standard shader can do this on real time, but I can't really understand what my code lacking

[meager pelican]

IDK off the top of my head. But if you're getting that quad, I'm thinking that some other condition needs that clip function call. Or "dither" doesn't have the value you think it does at the time.

You could try returning something else, like 0 instead of clip, but I"d assume 0 is the default.
GTG....work calls. 😉

[vale spruce]

damn... I'm dumb

_Cutoff ("Alpha Cutoff", Range(0, 1)) = 0.5
I forgot I changed my _Cutoff name

took me 2 hours I want to cri

[vale spruce]

is it possible to clamp the slider between 0 and 1 but if I want to I can still type like value of 5 manually?

[barren stratus]

Hey I have a question

so im trying to make a blend slider to slide between textures and Im having a bit of a problem

I tried making a frenel effect with the texture that'll be slided in and for now defaulted it to a normal color

for the slide in I used a Uv node and split it to take out it's Y axis value to perform the slide along the Y axis then connected that to the opacity of the blend node

it's having weird blending outputs. the blend mode is set to Overwrite so I have no idea why that is

I tried using normalise but it's still not working right

can I send a picture because I feel like that'll better explain it

[barren stratus]

That shouldn't be possible unless u create a script of some sort to handle the clamping and expand the clamping range if certain conditions are met

[barren stratus]

2 textures

[vale spruce]

I see... I guess I will give up using slider

[vale spruce]

better send picture I guess, I don't don't really understand your question. so like you want to blend 2 albedo texture then give the material a fresnel effect?

[barren stratus]

yh

one sec

i want the new texture to slide in like it's showing on the multiply and normalize nodes

but it's not

im currently just using the color blue to represent the texture

[vale spruce]

I don't really understand what you mean by slide in

[barren stratus]

Ive nv used blend node, literally just heard about it and wanted to play around with it

[barren stratus]

nvm then

tnks though, I'lljust learn something else

I think I'll wait for cyan to come on or something because thinking about the things I could do with that, my desire to learn it is peak rn

[amber saffron]

The terms "texture to slide" are pretty hard to understand :/

[vale spruce]

is there no way to force gpu instancing to take more than 2 pass?? :/

[hollow wolf]

does creating heartrate using shader worth the effort (performance wise) than creating it using particle?

[rare wren]

Depends on the exact shader and particles used.
I expect shader graph to be faster, since everything can render onto 1 quad, instead of multiple quads for particles. This saves vertex shading, and batching, in exchange for some higher pixel shader costs

[hollow wolf]

i see, but, isnt it like, heavy math if i want to create this shader

guess i need to learn math again lol

[rare wren]

Shaders if written properly are very efficient.
I don't think a heartbeat will be too demanding

But for now I'd say pick which is better for now. Unless you do something crazy both won't kill performance probably

[hollow wolf]

thanks for the info

[reef mango]

guys im really struggling to figure this out and i feel like im losing my marbles --------> https://forum.unity.com/threads/how-do-you-find-the-lowest-value-of-a-texture-sampled-at-different-locations.1465952/

[amber saffron]

What does it mean that the greatestHeight gives different values for different index ?

[reef mango]

thats the million dollar question. it makes no sense. how can greatestHeight not be the same number for every value of index?

[amber saffron]

I'm trying to understand how you know that it is not the same number.
Since for each index you sample a new heightmap value, if greatestHeight was updated before, it would make sense that it is different for each iteration of the loop

[reef mango]

let me clarify. when i write the output to a texture then i get different values at the different uvs passed in _ArrayCoordinatesOfSplats

[amber saffron]

Well, you are sampling the heightmap at different UVs for each value of the array, so this is not suprising ?

Oh, or maybe you were expecting that greatestHeightAtSpotLocations is a common variable shared between all the pixels ?

[reef mango]

yes. isnt it common to all index values?

[woeful breach]

does this look like clouds or more like water ?

[hollow wolf]

for me it's looks like cloud, simply because of the color and how it didnt transparance enought for a water

i dont know about the view when you near to it tho

[amber saffron]

In the context of the code you've showed in the forum post, it should be common for all "index values" when executing the loop.

[woeful breach]

good, because it is a cloud but a lot of people say it looks like water and heres a picture from a bit closer

[hollow wolf]

yeah, now it's really looks like cloud

[amber saffron]

What I REALLY don't understand is how you know that greatestHeightAtSpotLocations changes for each index, since you can really read the value only after the loop executed.
And how do you do to debug it in an output texture ?

[reef mango]

i write 'greatestHeightAtSpotLocations ' to a rendertexture after the loop executed

[amber saffron]

And so, the RT is not a solid color, the value varies for each pixel, even though the array coordinates are constant ?

[reef mango]

i see now something strange is happening when i blit the RT to a temporary RT and then that temporay one back to the RT

[amber saffron]

?

[reef mango]

when i draw red pixels to the RT and then blit that RT to a temp and from the temp back to the RT those pixels arent red anymore. They're grayscale.

[tight phoenix]

How do I get world space values that will allow me to lerp between top and bottom in world space from its bounding box in shader?

the space I thought I needed it in proved to be wrong :/ so now im back to square zero

i still can't wrap my head around the fact that min and max are not points in space and I can't just lerp between the two points in space

is there some way I could just write a subgraph to MAKE them points in space so that I can do this

because every single time I go to use them, I need them to be points in space I can lerp between but they arent so I cant

i can feel my distress already rachetting up because I need this done, I need this solved, I need it solved yesterday but I am no closer to a sollution and I can't do it on my own but I need to get it done so I am a failure who has to rely on others to help me and if I can't get help, im fucked

[reef mango]

@Remy it looks like it has to do with color formats

should i srgb on any of my texutres or must i stick with UNORM?

[amber saffron]

It depends on you heightmaps format, you should use the same

[amber saffron]

So, vertical gradient based on min and max bounds ?
result = (worldPosition.y - min.y) / (max.y - min.y)

[amber saffron]

Min and Max are points in space. But they are constant values for evey pixel of you object. So if you lerp between them with a constant mask value, you still get a constant value.

[tight phoenix]

I tried that, it didnt work

this didnt return a value between 0 and 1 from top to bottom, the masking didnt 'scale' it was still linear

[reef mango]

@amber saffron when i blit the source heightmap to the RT then the colors display correctly. But when i blit from RT to temp and the from temp to RT all the colors are grayscale.

[amber saffron]

Really ? It should (unless the object is rotated, then we gent to the local/world bounds issue again). How does it look in the scene ?

[tight phoenix]

the object is rotated/

I want world bounds this time, I was wrong yesterday, local doesnt work atll for what I need

[amber saffron]

Then the min is probably a bit lower that the lowest vertex of the object, and the max is probably a bit higher ... and this is expected

[tight phoenix]

I spent the entire day wasting everyones time and being a failure and ultimately the result wasnt even what I needed because as usual I am too stupid to even know what I need b

[reef mango]

@amber saffron i use R8G8B8A8_UNorm on all RTs and UnityEngine.Experimental.Rendering.GraphicsFormat.R8G8B8A8_UNorm on the temp RT

[tight phoenix]

I am in crisis and I cant handle this but I HAVE to immerse myself in this agonizing acid because its my job

sorry

[amber saffron]

Weird :/

[tight phoenix]

i should shut up

i cant handle being a failure but game dev every 15 minutes you're failing at something

[amber saffron]

Are you expecting the gradient to go from black to white from the lowest vertex/pixel to the highest one, independtly of the rotation of the object ?

[tight phoenix]

Yes

I want a float value where 0 stepped is fully masked out top to bottom, and 1 is fully masked in top to bottom

[amber saffron]

Won't happen then unless you recalculate the bounds yourself each time the object is rotated then.

[tight phoenix]

no matter what size or rotation the mesh is at

okay so I can pass in that bounds again from the script, except per frame instead of once

[amber saffron]

Not enough

[tight phoenix]

what why?

[amber saffron]

Let me try to sum up this :

• You want the world minimum and maximum Y of the mesh, whatever the rotation and scale
• You can not use either the local nor the world bounds, as :
  • The local bounds will outshoot minimum and maximum when the object is rotated
  • The world bounds mush encapsulate the local bounds, and will also always outshoot

So you need to calculate world bounds (or min and max Y at least) each time the object is scaled/rotated, by taking in acount all the vertices world position

And this can prove quite costly if you have a lot of objects with complex meshes

[tight phoenix]

😰

I don't understand why its so difficult or costly, its frustrating because in real life its incredibly simple to say 'here is the top and bottom of something'

like I am not trying to write code for it to identify what is or isnt a bird in a picture, I just want the top and bottom

trying not to panic

[amber saffron]

Because cpu and gpus don't work like you brain, and realtime rendering needs as fast as possible (so as simple as possible) calculations.
The mesh bounds is calculated once, when the mesh is imported.
It then can be summed up to 8 points : the corners of the bouding box
The world bounds is calculated in realtime by quickly calculating a box aligned with the world coordinates that includes those 8 points.

This is very fast, compared to having a more precise bounds for meshes that coulds contain million of vertices