archived-shaders

[strange basalt]

so multiple ways to do it via code, but without code more or less you make render texture assets

then on a camera you can tell it to render into that texture instead of the screen

[stiff zenith]

Does that mean I'd need 1 camera per car displayed?

[strange basalt]

it could it depends on your needs

are the cars static, do they change in any way?

[stiff zenith]

Yes

[strange basalt]

so yes they change or yes they are static?

[stiff zenith]

Their color change dynamicly

[strange basalt]

while they are in icon form?

[stiff zenith]

Yes

[strange basalt]

yeah so would need 1 render texture and 1 camera per

if they were static you could do it with 1 of each, then just create regular textures at runtime to use later

[stiff zenith]

I got another idea, as they are rendered with an orthographic camera, can't I use 1 camera to get the texture for them all?

[strange basalt]

that is possible as well

[stiff zenith]

Nice

I'll look into that

Thanks a lot!

[strange basalt]

to use them in the UI you will need the RawImage component instead of a regular image componet

[stiff zenith]

Oh

Ok

[strange basalt]

it has UV rect so even if you rendered them into 1 texture, its still easy to have different UI compnents only show 1 part of the texture

if that makes sense?

[stiff zenith]

Yes it does

Thanks!

[strange basalt]

that should make aligning it in the UI easier, 1 texture with all cars, but 1 RawImage component for each car you want to display in the UI with different UVRect settings

can use layers and culling settings, to have the camera capturing it only see the cars you want and not other thigns in the scene

[ionic python]

i wanted to do that actually!
i just dont know how to lmao

[strange basalt]

i dont have unity open right now, but iirc there is a node called SphereMask

you tell it where the center is, then you feed in the world coords as well

then you can adjust the radious

[ionic python]

I was thinking of making it so like an explosion happens when the scythe gets smashed down

well shit you got me curious now lmao

its 2 am

i dont get how you'd do that though

oh

[strange basalt]

something like this going into your lerp

[ionic python]

for time?

[strange basalt]

well to use time one sec

ignore the errors i dont have it all hooked up

max radius you would just have to figure out what value you need

same for Center since 0, 0, 0 would be center of the map, so you would want to set that to where teh center of where the effect should be

but with this you could adjust combattime and the effect would spread from 1 point outwards

[broken vigil]

I just feel like venting for a quick second, but... God, I think the thing I hate the absolute most in Unity out of everything is how much it anti-supports letting people access and control lighting data in custom shadergraph shaders.

I've been fighting with making a Toon Shader for so long, and it feels so annoying, because all the lighting data is right there, it's not not possible for me to easily modify it using stuff like ramps and color modification.

Is there ANY solution in 2024, when all I want is a node that outputs the full suite of URP lighting, as if an object were lit, but ALSO allows me to modify that output for stuff like Toon Shading and Lighting Ramps?

[ionic python]

what do I hook the sphere mask into tho? @strange basalt

[strange basalt]

you can just write hlsl as you need, but yes 100% i wish it would let people write there own master nodes, then use shader graph to drive the inputs to it

[ionic python]

wouldnt i need to multiply or smth?

[strange basalt]

the lerp that goes between your textures and flat version of things, to its t input

[broken vigil]

tbh, I don't consider "just write hlsl" to be an answer either. with how godawful Visual Studio's hlsl support is, and how bad the documentation for Unity's hlsl implementation is, it's nearly impossible to figure out ANYTHING about it. Even ignoring how much of the documentation you can easily find is literally outdated or does not work.

My best attempts at doing things in HLSL literally just ended with me asking ChatGPT to write it, after I went through multiple websites worth of information where literally none of it works.

[ionic python]

doesnt quite look right @strange basalt

[strange basalt]

to each there own, i would not want to might my own lighting model in shadergraph even if it did let me do it. but see lots of benefit of desing my own master node with my stylized shading then having shader graph as a artists friendly way to play with it

[ionic python]

nvm me stupid

i also want to have a transparent sphere which shows how its happening

[strange basalt]

cant really do that in the shader, so would just have that as a seperate effect

a sphere with its own shader scaling up

[ionic python]

yea thats what im doing

[ionic python]

@strange basalt

[strange basalt]

good job, is coming along

most good effects are like this, really takes multiple things comiong togeather to make it

[ionic python]

ill give the sphere some distortion effect to really sell it

add some noise to it to make it less boring

[strange basalt]

bit of screen shake, or messing with the camera FOV too

[ionic python]

something did get fucked up with my materials tho

[ionic python]

there is screen shake for the scythe slam

[ionic python]

@strange basalt

this is what the gameplay looks like...kinda... lots to improve on

compression killed the vid

[strange basalt]

still looks pretty cool already

[ionic python]

I need to work on a lot of camera shit

its killing the game pretty hard

plus I need to learn how to make actual shaders

and not just these random effects

[strange basalt]

camera is one of the harder things to polish

[ionic python]

I only know barebones hlsl

I was thinking of restarting from scratch

[strange basalt]

it will come with time, like programming doing shader is very much just trying to break down a problem as much as possible

and seeing things you like elsewhere and trying to reason aobut how

[ionic python]

lots of my systems are really badly written

and a lot of things that id love with projectile manipulation just isnt possible without rewriting a lot of things

plus theres DOTS which would be extremely useful for optimization

[strange basalt]

are you hitting the need for such optmizations though

it is a very different way to approach thigns and not as free form or prototyping friendly

[ionic python]

my laptop has some issue where at a random time it decides to shit itself and cause every game to run below 10 fps

[strange basalt]

so if the performance drop feels very random

high chance its garbage collection

[ionic python]

its not even just the game

[strange basalt]

but you can use a profiler to figure out what it is

[ionic python]

I can run elden ring at 110 fps

but sometimes it gets so bad that a built empty scene runs at 40 fps

[strange basalt]

yeah something sounds pretty off there if a empty scene is doing 40fps

[ionic python]

i was thinking of a reinstall

like factory resetting

[strange basalt]

is this just in editor frame rate?

[ionic python]

no, I built an empty scene

[strange basalt]

editor has a lot of overhead compared to a build

[ionic python]

other times the game runs at a solid 120 fps (i capped it)

but yeah

anyhow

[strange basalt]

does it effect non unity things?

[ionic python]

I just don't know how to go about restarting from scratch

[strange basalt]

or perhaps the same drops on other unity games

[ionic python]

yeah

[strange basalt]

ah so something weird on the system maybe thermal throttling

[ionic python]

elden ring drops down to 5 fps

not thermal throttling

the cpu is cold

sometimes it starts burning my hands and it still runs at perfect fps

feels like I have a fucking crypto miner on my pc

[pure vapor]

would it be possible to get the (probably wording this wrong) "light data" from like spotlights n' stuff

cuz i have an unlit shader and i wanna attempt to add support for lights like that

kinda dumb buuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuut

[ember summit]

Hey there! In shadergraph, you see a standard chroma key result on a video render texture using a Color Mask node. note the pure green mask color selected in the inspector.

In the 2nd image, note the high range value and that the avatar's eys and hair contain colors no where near the desired mask color, yet those are turning transparent while there is still a green edge in the resulting image. Something wrong with the Color mask Range calculations if the green is still in there at the range value shown.

What gives? Do I need to pipe this through multiple color masks so I can chose multiple shades of green?

[tardy sail]

Hiya - I'm trying to put a renderMesh on a quad in my 2d game. So far I:

1. Have a 3d camera that outputs the rendermesh
2. Created a material out of the renderMesh
3. Have a quad in the scene
4. Assigned the material to the quad
5. Lit the scene

The result is a material that looks good in the inspector - I can see the character.
The quad is completely transparent, though - I see nothing.
When I add "Lit" as a second material to the quad, I can see the character, but the whole quad takes on the background color of the camera (even though the camera background is set to 0 alpha)

I think from some googling that I need to edit the shader (which happens to be Sprite-Lit-Default) to support the alpha from the camera? But I'm wayyyy out of my depth looking at this shader file

This: https://discussions.unity.com/t/render-texture-on-2d-sprite/677593/2 is a very brief pointer of what I want to do, but the person there leaves it at "*still need to use some transparent material shader i think"

[tardy sail]

nvm - figured it out through sheer luck by setting the shader to unlit-transparent-cutout

[weary dragon]

is there a way to not make the shader graph as bright? (left is shader material, right is skybox.)

this is the shader graph inside btw

[shell phoenix]

yeah just multiply the last node with any float variable.
The shader graph you posted already has such a node but it is connected to emission instead of base color. (They are basically the same but emission ins't used here)

Does anyone know if subshader dropdowns behave like keywords?
I want to compile different code in hlsl based on which dropdown option is used but I don't see what refernce it uses.

[waxen merlin]

Compute shader optimisation question. I know that if you have a compute shader where for each pixel you need to sample the 8 neighbouring pixels, then it is better to do those texture samples first (once for each pixel) and store them in group shared memory, add a barrier and then process them from the group shared memory.... is this also the same if it's not a texture, but a compute buffer? Is querying the array first and putting the values into group shared memory also worth it? Thanks.

[obsidian wasp]

Anyone know why a dither node is crapping out at certain resolutions / window sizes?

[quaint grotto]

i have water shader which uses the depth buffer and ive set the shader to render both front and back faces. but for some reason the water depth visuals only works on one side of my plane mesh, why is this?

it looks like glass until i rotate it

shouldn't it work for both sides if i have both front and back face rendering on?

it works fine if i duplicate and 180 the mesh but i would rather not

[tacit parcel]

maybe because the normal facing backward on the back face.

[quaint grotto]

right but im not using the normal in my depth calculation

im using depth buffer to world position and fragment world position

calculating the difference and then converting to 0 and 1 range

unless the depth buffer itself uses the normals in a way i do not know about

let me just do a quick test with a grey output to see if thats what it is

well damn

its the depth buffer but i dont understand why

[quaint grotto]

this has to be bug surely? i have front and back face rendering on and im putting scene depth straight into color output, why would it go black when i rotate the plane

[tacit parcel]

could it be negative? try absoluting it

[quaint grotto]

yeh tried - its still black

seems the values are 0

its as if they zero it out based on the mesh normal perhaps but thats an odd way to calculate depth

@tacit parceldont suppose you could try replicate to check ?this is the entire graph

if it doesn't happen for you then ive broken something lol

[weary dragon]

which is the last node and what do you mean by multiplying it? (im kind of new to these shader graph things)

is this how i do it or am i doing something wrong

[weary dragon]

im trying to make it the same as the skybox

i have also seen that it tints the sun and some clouds if they are bright enough.

[grizzled bolt]

Because your Game window has a non-uniform scale for previewing purposes
It has no adverse effect on anyone natively running your game at those resolutions

[shell phoenix]

Yes that is the right setup. Now you can adjust the brightness via the the "emission tint" variable in the material inspector.

[shell phoenix]

They ads probably tinted because the emission tinted color isn't pure white

[stray orbit]

Hello, i am trying to offset my vertices a little to achieve a sort of pulsating motion. What i want to achieve is that the object that the material is applied to kind of shrinks and grows very slightly. What i have right now is a sine wave time value multiplied by a factor and added to the vertex position. The object fully offsets in one direction which is not yet the desired result. I would like to know how i can get the direction from the center of the object to use it in the calculation and offset the vertices correctly

[kind juniper]

Get the normal of the vertex, multiply it by you offset and add to the position.

[stray orbit]

This seems good, do you know by any chance if there is a way to do it through the origin of the object? i think that might be even better for my usecase

basically move away from or towards the origin

[kind juniper]

Could try subtracting the vertex position from the object position and using that as the direction.

[stray orbit]

i'll see how that will look thanks

[kind juniper]

Or the other way around actually

[regal stag]

Add 1 after Multiply, then Multiply with Position instead of Add

[stray orbit]

This happens because the faces are not connected right?

[dim yoke]

Yes, smooth shaded mesh wouldn't give gaps

[waxen ridge]

Hello,

any ideas why this could be happening?
the first model is something from an asset, the second one is after importing the fbx to blender and then exporting it.

ALL model import settings are identical.

The first one has an outline, and the second doesn't, using unity's toon shader.

[sullen loom]

performance question for shadergraph. Is it expensive/bad-practice to use The Length node? I know in c# land doing those sqr root calculations is sometimes touted as a "no no" but not sure what kind of optimizations go on in shaderland. I didn't see a sqrDistance node or anything so is it "ok" to use Length?

[low lichen]

I think it's about as expensive. If you can avoid it, it's better to do so. To calculate the square length of a vector, you just need to get the Dot product of it, where both inputs are the same vector.

[sullen loom]

Ok great to know thanks a lot!

[waxen ridge]

aparently i was not exporting tangent space normals

[dim yoke]

Avoiding expensive calculations is always preferable when possible but GPUs are stupid fast. As an example, the SimpleNoise function does 6 pows, 12 sins and dozens of additions/multiplications for each sample. Even then many use those procedural nodes (often even many of them at the same time) in pixel shaders and it will be totally fine on modern PCs. If you target low end mobile devices or something like that, then micro-optimization might be fine but on modern PCs calling sqrt "no no" is really silly on C# side either. Use what you need, no more no less, try to optimize only in very specific situations

[sullen loom]

yeah I agree it's not always as big of a deal as people make it out to be on the C# side

[dim yoke]

It's usually not as big of deal. For example, when I need to do circular shapes in shaders, I don't even care if there's way to optimize it (because usually there's not really), using Length is totally fine. Redundant calculations should always be avoided but you would be surprised how capable GPUs are, there's no need to waste hours trying to avoid operations that won't really affect the performance anyway in the bigger picture

[craggy hare]

I am trying to recreate Physics based sky from this[1] git hub repo that is based on this[2] article. Repo is written in C++ and OpenGl. I rewritten all of the code to be HLSL and the problem is that my skybox just renders black. I tried tweaking some values and equations, but it had no effect. I am quite new to shaders, but I found topic of Physics based sky to be very interesting. I would like to hear any suggestions how to fix my shader.
[1] https://github.com/newpolaris/Atmospheric-Scattering/tree/3a3b96a136e9dd703ca037cfbca071749a183f1d
[2] https://www.scratchapixel.com/lessons/procedural-generation-virtual-worlds/simulating-sky/simulating-colors-of-the-sky.html
My shader code: https://pastebin.com/crNYR48X

[solar current]

Is it possible to use a custom mesh in place of the default Unity skybox sphere?

[snow carbon]

yo I want to make my toon shader cast toon shadows on other objects that have the same toon shader enabled, does anyone know how I can do that?

[solar current]

What Rendering Pipeline are you using? In URP you can make it so only objects on the same Rendering Layer will cast and receive to one another. These shadows are saved to a Shadowmap texture that each shader samples from based on its layer (if Shadow Layers is enabled), so you might be able to override the Layer used by the shader in the shader's Forward pass

[dim yoke]

As far as I know it's not possible. Is there a specific reason why you would need it?

[waxen grove]

Hi, can someone help me project a moon texture onto the skybox? I'm using a modified built-in procedural skybox

I'm trying to make it more like the one of the right, but it's the one on the left

[waxen grove]

OK, I did manage to do it, but there's something wrong. Whenever I go towards the edge of the, it does flattens the projection, see right. How can I fix this?

Here's how I calculated the UVs
half2 _moonUV = (((_WorldSpaceLightPos0.xyz - ray) / _MoonSize) * float2(-0.5, 0.5)) - float2(0.5, 0.5);

Also, this is how I did the texture: col += tex2D(_MoonDetail, _moonUV)

[mental vale]

Hello guys, I have an error about the Shader Graph Package. I don't know how to fix this I tried to add this "com.unity.shadergraph": "14.0.11" to the manifest.json file but it's not working.

[kind juniper]

Did you actually install it via the package manager?

[mental vale]

yes

[kind juniper]

I wonder if it's a valid version. Try deleting it and installing the package again.

[mental vale]

alright

[kind juniper]

Also, what unity version are you using?

[mental vale]

2022.3.0

[waxen grove]

Hey dlich, can you help me please with this?

[kind juniper]

Maybe update to the latest 2022.3 as well.
In the shader graph change log they specify the unity min versions for some reason.

[kind juniper]

Whenever you go to the edge of what?

[mental vale]

someone suggested me to change the Unity version

[waxen grove]

Like, when it's close to the x axis
I know it has to do with how the UV's on the skybox are

[kind juniper]

And what are all the variables in the calculation? There's a lot of context missing to be able to help you.

[waxen grove]

I'm using the built-in procedural skybox

I've just modified it to have the moon, clouds and stars

[mental vale]

alright it's working

[kind juniper]

Well, that's a lot of changes. And it also implies I need to know how the built-in procedural skybox shader works, which I don't.@waxen grove

[waxen grove]

Oh

[kind juniper]

Maybe share the whole shader. I'm not sure I'll be able to help though.

[waxen grove]

Here's what I have so far, what you are looking for is on lines 436 (Drawing the moon) and line 376 (Calculating the UVs)

[kind juniper]

!code

[echo moat]

Posting code

📃 Large Code Blocks

Use links to services like:
https://gdl.space/, https://paste.ofcode.org/, https://hatebin.com/, https://paste.myst.rs/, https://hastebin.com/

📃 Inline Code

Surround code with three backquotes. Not quotation marks.
To format as C#, add cs to the first line:
```cs
// Your code here
```
Add a comment with a line number if there is an error message.

[kind juniper]

Also, can you share a link to the original shader?

[waxen grove]

Here's the original link (It is literally just a modified version of the default skybox)

https://github.com/TwoTailsGames/Unity-Built-in-Shaders/blob/master/DefaultResourcesExtra/Skybox-Procedural.shader

Also, I can't do the code blocks because the script is too long

Here's a more generalized view about my problem- See where it collapses on itself? I want to avoid that when the sun is facing that direction

[kind juniper]

I've read the original shader a bit, but it doesn't make sense to me. I don't think I would be able to comprehend it without spending a considerable amount of time reading though it and testing.
Perhaps someone else would be able to help you. Just share your shader correctly.

[waxen grove]

I would, but I literally can't... it's too big to fit the whole shader, and I've already included the parts that do matter

[kind juniper]

I mean share it via paste site. Or upload to GitHub or something.
!code

[echo moat]

Posting code

📃 Large Code Blocks

Use links to services like:
https://gdl.space/, https://paste.ofcode.org/, https://hatebin.com/, https://paste.myst.rs/, https://hastebin.com/

📃 Inline Code

Surround code with three backquotes. Not quotation marks.
To format as C#, add cs to the first line:
```cs
// Your code here
```
Add a comment with a line number if there is an error message.

[waxen grove]

Oh, I see I see

🤦 I idio

https://hastebin.com/share/ejarotokib.cpp
Here you go

[mental vale]

Guys, I'm following this tutorial https://youtu.be/_RbxRYAWl3g?si=OQgn74ylAzgJkvDU and I installed the HDRP Package. But when I create the Sky and Fog Volume the inspector is not the same as the video, I mean the Component is not the same so I can't follow the video check on 0:46

[weary dragon]

it was a light gray and i changed it to white, but it still doesnt seem the same. its fixed the weird yellow bit, yet its made it more bright.

[zenith jetty]

(shader graph) I'm getting a significant performance hit when using a sample texture 2d LOD in the vertex shader on mobile, there's also no performance difference between using 1 or 3 sampler nodes, it's just "on or off", anyone knows how to avoid that except not use texture samplers?

[dim yoke]

How large is the texture? My wild guess would be that it might have to do with the memory overhead and not the performance if the sample count doesn’t even matter

[zenith jetty]

It's just 64x64

[dim yoke]

Nvm then

[sonic axle]

Hey is there any good unity shader graph tutorials?

[stiff zenith]

Hey, how can I make a shader that ignores mip maps for a texture?
Here's the base for the shader: https://github.com/TwoTailsGames/Unity-Built-in-Shaders/blob/master/DefaultResourcesExtra/Standard.shader

[grand jolt]

Hi, very beginner question, I'm trying to find sprite-lit-default but I can't see it, does anyone know how to?

[waxen ridge]

That sounds like a shader and not a material. Create a new material, and add the sprite-lit-default shader to it. (I don't know much about 2d)

[grand jolt]

Ok I fixed it... it was just that I didnt make the proyect an universal, I got confused with the name, then I installed an Universal 2d pack on the pack manager and fix it (just in case someone has this problem)

[kind juniper]

You'll need to modify the fragment/pixel shader to always sample the desired mip level(i.e. 0 which is the same as the original texture).

Or just disable mipmaps generation for the texture.

[stiff zenith]

Ok thanks, I'll try to do that

[stiff zenith]

I still want it to have mipmaps when used in other places so I'll take the first option

[kind juniper]

Might be easier to have 2 texture variants then. One with mipmaps and one without. It's a bit of an overkill to modify a shader just for that.

[stiff zenith]

Ok

Thanks for your help!

[kind juniper]

Had a typo in the message that made it sound ambiguous, so check it again to make sure you got it right.

[stiff zenith]

I understood it right, no worry. Thanks

[cyan maple]

Is there a way to make a billboard shader ignore certain rotation axis of the camera?

https://cdn.discordapp.com/attachments/288230143686606858/1276771179855741001/Desktop_2024.08.24_-_14.51.38.02.mp4?ex=66cabd52&is=66c96bd2&hm=dba69dad9e61db0416cec31b9836d7f43cf94f12c3819a141d7ecb3c77df82bb&

I'm trying a non-volumetric cloud solution and it works okay, I just want the billboards to ignore the camera's z rotation so the clouds don't rapidly rotate when the aircraft spins

https://cdn.discordapp.com/attachments/288230143686606858/1276773105792188516/image.png?ex=66cabf1d&is=66c96d9d&hm=154f5de144c9830e51279fbe9ff1761d3ef6d45e4ec65968b9a06de26ad4b7c0& setting it to a lit shader also doesn't seem to be working right :/

(the white clouds are just default lit shader, the blue ones are the billboard. The color should be the same)

[nova bear]

How do I pass vertex data to this shadergraph?

My code

it gets position and normal right

but materialType doesn't get correct

[near pasture]

I know shader graph, but I'm not a C# guy. Still I'll attempt to help 😅 If you're trying to pass something to the shader graph through a script, you can expose parameters in the graph and change them with the script.

[nova bear]

Thanks for the attempt!

I can expose stuff such as setting textures and so on. I need to pass per vertex data

[near pasture]

What kind of per vertex data do you need to pass?

[nova bear]

this,

[regal stag]

Afaik shadergraph generates specific vertex inputs (position, normal, tangent, texcoords should all work fine) - but doesn't expect blendweights. If unused, you could use texcoord1 though. Which you'd obtain in graph using UV node (no need for custom interpolators either)

[nova bear]

And this is struct (of array) I wrote to pass

[regal stag]

Alternatively could try passing an array/buffer to the material, which you could then index with the VertexID

[nova bear]

I would honestly just prefer to pass array of vertex data instead

I tried texcoord1 and as well attempting to just use texcoord0 with 3 dimensions to fit the size of vertex

it just doesn't work

I don't know what's wrong

[nova bear]

I feel like this structure is meant for script based shader coding

I literally cannot find any resources on passing data for "Custom Interpolator Node"

[regal stag]

It should work in both cases, but you don't need to do anything in the graph to support it - don't add a custom interpolator. Just use the UV node as you would usually - shader graph handles passing it between vertex and fragment stages for you behind the scenes.

[nova bear]

So this is the new layout as you suggested

it still doesn't work all UV and MaterialType becomes 1

when I populate per vertex data I set these to 0 to see if 0 gets passed

this is rendering UV part as color and as you can see it is all yellow

which means, 0 was never passed

[regal stag]

Are you using the UV node or your custom interpolator still?

[nova bear]

The red box ones came with the shader graph

I added the last two as interpolator

[near pasture]

I'm not sure why you're using custom interpolators at all. Are you trying to have "Material Type" driving something in your shader?

[nova bear]

I dont think I am using custom interpolator

[regal stag]

Remove them and just try using the UV node in the fragment stage. I don't know how to make this any clearer

[nova bear]

I think the way I am trying to use the custom interpolation is inorrect

[near pasture]

Aren't the two custom things there called "custom interpolators"?

[nova bear]

yes

but when I pass data to vertex

this is how I do it

[regal stag]

Custom interpolators are for calculating something in the vertex stage and passing to the fragment. They are not for passing data from the mesh

[nova bear]

and I dont think this is how you pass data to custom interpolators

I want per vertex data to be passed not per mesh

[regal stag]

Do this

Is the result all black?

[nova bear]

yes

let me change the values as well

to see if I can pass actual numbers

Commented out the 0s

yes it has been passed

[regal stag]

Nice. To access the materialType in shader you'd use another UV node but set the channel to "UV1" instead of "UV0". Can use Split (R output) or Swizzle (with "r" or "x" in field) nodes to convert down to a float

[nova bear]

@regal stag Bunch thanks

https://www.youtube.com/watch?v=LBX7MUhvsSw

Sweet

❤️

[coral hinge]

how do i do something like this in unity

[stiff zenith]

Hey, is there any way I can add a noise map that controls the alpha of a texture using a shader and make it so that if alpha is already 0, it stays at 0? I'd like to make my tags on the road feels like they're not freshly painted

[naive elk]

I was hoping to find someone decent with the shader graph to help me mask a grid shader/ EX Show the grid around an object/mouse then dfade as it gets further from the object

[kind juniper]

A cylindrical mesh + a transparent material + a texture that has alpha fading towards the edges.🤷‍♂️

[kind juniper]

Store the noise in a separate texture and multiply it with the original color output.

[heavy sierra]

Hi guys,
Does someone know how can set render targets, what are the steps? I have 3 render targets to set. See more in thread

[heavy sierra]

Shader falls back to default/diffuse

[haughty bear]

Hello everyone! Could you please advise me on how to remove the seam in a shader? The shader isn't a pattern, so I need to somehow blend it, but I'm not quite sure how. The object is a cylinder, I've tried different UV unwrappings but came to the conclusion that it's a shader seam. I followed a YouTube tutorial to create this shader, so I'm not sure how to hide this joint. Thank you in advance.

[regal stag]

Use a texture property with the reference _MainTex

[grand jolt]

is there a way to extract the camera position and rotation from a model view matrix or a projection matrix?

[regal stag]

One method would be to use a "seamless noise texture" rather than the procedural noise nodes, and an integer tiling value. (Also assuming the cylinder mesh takes up the full 0-1 x axis on the UVs)
May be able to find one online or search for a method of generating one yourself

[glossy quail]

I have a quick question, I am using Instanced meshes, but if I bake lighting they are not included in the light built, do I need a special shader which also works for instanced meshes?

[regal stag]

_WorldSpaceCameraPos should give you the camera pos. I think UNITY_MATRIX_I_V._m03_m13_m23 might also be the same.
As for "rotation", that isn't as useful in shaders, you'd usually just use the matrix to apply the transformation instead of extracting that

[grand jolt]

no like im just curious is there a way

[neat orchid]

I'm stupid

why does this not work

I copied the texture blend section of the polybrush shader of the albedo

I'm trying to make one for normals but idk what I'm doing

[torpid sundial]

Hello, I am very inept when it comes to shader graph in Unity, and I was wondering how I would go about making this laser shader have a waving effect. I'm sure it's something simple to modify the UV and use Time node but I just can't get it to work... 😅

[sleek radish]

I keep getting this error:

Shader error in 'Custom/CubeShader': unrecognized identifier 'VaryingsMeshToPS' at (HDRP)/Library/PackageCache/com.unity.render-pipelines.high-definition@14.0.10/Runtime/RenderPipeline/ShaderPass/VertMesh.hlsl(3) (on d3d11)

Compiling Subshader: 0, Pass: ForwardLit, Fragment program with <no keywords>
Platform defines: SHADER_API_DESKTOP UNITY_ENABLE_DETAIL_NORMALMAP UNITY_ENABLE_REFLECTION_BUFFERS UNITY_LIGHTMAP_FULL_HDR UNITY_LIGHT_PROBE_PROXY_VOLUME UNITY_PBS_USE_BRDF1 UNITY_SPECCUBE_BLENDING UNITY_SPECCUBE_BOX_PROJECTION UNITY_USE_DITHER_MASK_FOR_ALPHABLENDED_SHADOWS
Disabled keywords: SHADER_API_GLES30 UNITY_ASTC_NORMALMAP_ENCODING UNITY_COLORSPACE_GAMMA UNITY_FRAMEBUFFER_FETCH_AVAILABLE UNITY_HALF_PRECISION_FRAGMENT_SHADER_REGISTERS UNITY_HARDWARE_TIER1 UNITY_HARDWARE_TIER2 UNITY_HARDWARE_TIER3 UNITY_LIGHTMAP_DLDR_ENCODING UNITY_LIGHTMAP_RGBM_ENCODING UNITY_METAL_SHADOWS_USE_POINT_FILTERING UNITY_NO_DXT5nm UNITY_NO_FULL_STANDARD_SHADER UNITY_NO_SCREENSPACE_SHADOWS UNITY_PBS_USE_BRDF2 UNITY_PBS_USE_BRDF3 UNITY_PRETRANSFORM_TO_DISPLAY_ORIENTATION UNITY_UNIFIED_SHADER_PRECISION_MODEL UNITY_VIRTUAL_TEXTURING

Here is my shader: https://gdl.space/anenupekov.cs

HDRP btw

[cosmic prairie]

hook Time into a Sine function and add it to UV's y component

[karmic hatch]

What do you mean by waving? Make it wiggly?

[torpid sundial]

yes, wiggly, like a sine wave

[torpid sundial]

tried this, but it just moves the whole thing up and down, not quite what i was looking for

[tacit parcel]

Try inputting uv.x instead of time

[stray orbit]

Hello, i would like to scale my UV's back to the actual texture size in shader graph.
I am following a video where it says uv's are normalised between 0 and 1.
They use
data.uv * _MainTex_TexelSize.zw;
Do i just use the texture size node and put the texel width and texel height value into a vector4 to multiply or the regular width and height?

ah i just use a vector2

[stoic oasis]

How do I include 2 colors in a normal map? For example, lets say I have a brick normal map, and I want the bricks to be red, but the crevixes to be black. How would I go about doing this?

[tacit parcel]

normal map contains surface normal direction, it cant contains more color, let alone 2 colors...
but you can add 'color' information into normal alpha channel, and parse it anyway you like (example, alpha > 0 is red, else it's black)

or, maybe you use the blue channel, but then you'll need to parse normal data manually

[mental bone]

Normally you would use a base texture aka albedo texture for the color.

[amber saffron]

You could also take inspiration of this blog post, where it packs normal + smoothness + color "gradient" in a single texture, that is then used to mixe between two constant colors : https://unity.com/fr/blog/engine-platform/experimenting-with-shader-graph-doing-more-with-less

[true meteor]

idk if its a right channel but how do i use roughness maps in unity?
I know that i have to invert values cuz unity uses smoothness and not roughness, and then put it in alpha channel of metallness map, but i dont think it works for me, anyone can help me?

also im using urp

[grizzled bolt]

If you're not making custom shaders, ask in #archived-urp
If you are making custom shaders, show your work

[tame topaz]

Please remove any offensive content from your bio, for this server, thanks.

!mute 520992992899760198 If you want to be childish be my guest. When you decide to grow up, you can return and use this server.

[echo moat]

realotoko was muted.

[stoic flint]

can someone help me make this vert billboard ?

        void vert (inout appdata_full v, out Input o) {

            #if defined(PIXELSNAP_ON)
            v.vertex = UnityPixelSnap (v.vertex);
            #endif
 
            float4 color = v.color;
 
            // waving
            float factor = (1 - _ShakeDisplacement - color.r) * 0.5;
            const float _WindSpeed  = (_ShakeWindspeed  + color.g );
            const float _WaveScale = _ShakeDisplacement;
            const float4 _waveXSize = float4(0.048, 0.06, 0.24, 0.096);
            const float4 _waveZSize = float4 (0.024, .08, 0.08, 0.2);
            const float4 waveSpeed = float4 (1.2, 2, 1.6, 4.8);
            float4 _waveXmove = float4(0.024, 0.04, -0.12, 0.096);
            float4 _waveZmove = float4 (0.006, .02, -0.02, 0.1);
           
            float3 worldPos = mul(unity_ObjectToWorld, v.vertex).xyz;

            float4 waves;
            waves = worldPos.x * _waveXSize;
            waves += worldPos.z * _waveZSize;
 
            waves += _Time.x * (1 - _ShakeTime * 2 - v.color.b ) * waveSpeed *_WindSpeed;
            waves = frac (waves);
            float4 s, c;
            FastSinCos (waves, s,c);
            float waveAmount = v.texcoord.y * (color.a + _ShakeBending);
            s *= waveAmount;
            s *= normalize (waveSpeed);
            s = s * s;
            float fade = dot (s, 1.3);
            s = s * s;
            float3 waveMove = float3 (0,0,0);
            waveMove.x = dot (s, _waveXmove);
            waveMove.z = dot (s, _waveZmove);
 
            v.vertex.xyz -= mul(unity_WorldToObject, waveMove).xyz;
 
            UNITY_INITIALIZE_OUTPUT(Input, o);
        }

[haughty bear]

Thank you, it really helped, although the result is not what was expected, but it fits. And how to open shaders that are "outdated"? I found a good shader that I need, but it doesn't work in my version of unit, how to fix it? Instead of a shader, I have a pink color. https://github.com/JPBotelho/Soap-Bubble

[regal stag]

If you have a new question please ask it to everyone rather than in a reply.
But surface shaders only work in Built-in RP. If working in other pipelines you would need to rewrite it manually, or find another example for your render pipeline.

[haughty bear]

Got it, thank you very much. I apologize for asking the question incorrectly.

[stiff zenith]

What shader should I use for a game with shadows/pixel lights but no reflections and no PBR, just opaque materials with metalic and smoothness? I feel like standard shader has too many features that I don't use and that may be slowing down rendering a bit

Something like #1263105452112351232

[grizzled bolt]

"Metallic and smoothness" are what PBR is, basically

[stiff zenith]

I meant no pbr texture

[grizzled bolt]

Whether it allows you to use a texture or not doesn't make a practical difference

[stiff zenith]

Oh

Ok

[grizzled bolt]

URP has "simple lit" shader and BiRP has "mobile" shaders which are cheaper

Shader features you don't use on a material get compiled out of that shader variant and so won't cost performance, at least on SRPs

PBR shaders are somewhat more expensive even with the same setting than a traditional non-pbr specular shader, afaik

[exotic steeple]

do you guys ever use SystemInfo.graphicsShaderLevel to determine if you can use a specific shader feature?
https://docs.unity3d.com/6000.0/Documentation/ScriptReference/SystemInfo-graphicsShaderLevel.html

[exotic steeple]

turns out all my phones return graphicsShaderLevel 50 lol ... all 8 of em

[stiff zenith]

Thanks

[spring mist]

that's quite literally the peak for most phones, 6 and 6.5 are outside of the phone capabilities currently

[hoary marsh]

what happened to my shader, why work like this(dividing screen in half) (i don't understand shaders in i watch video and make this because i want to try to make shader)

[regal stag]

Check settings on the renderer feature used to apply the shader to the screen

[hoary marsh]

Here ?

a and this shader is

ok i find problem in video author shows this work on blit but on my project when i switch to drawprocedural ad its works

[elfin bison]

Hello! I wanted to know if it's possible to create a fade effect in my shader using Shader Graph. I'm using a transparent material, and I think the issue might be related to alpha clipping. However, I've unchecked it in my Shader Graph, so I'm not really sure what's causing the problem. If anyone has a solution, I would really appreciate it! Thank you!

Here is my mask that i put in alpha !

[rare wren]

Alpha clip cannot fade. Why is it on?

[elfin bison]

ok thank a lot , i don't know i deactivate it but i still have the entry on my shader

[snow forge]

Hi all,

Just wondering if anyone had any ideas on how to achieve this kinda of effect with URP Shader Graph?

https://streamable.com/emspu2

I can get the texture to 'scroll', but not 'evolve' like in the video.

[stiff zenith]

You could use 3D noise and just change the z axis to make it "evolve"

[snow forge]

Forgive the ignorance, but how do you use 3D Perlin noise in Shader Graph?

[stiff zenith]

I don't know

[snow forge]

Ah. lol.

[stiff zenith]

https://discussions.unity.com/t/i-made-some-custom-nodes-for-3d-noise-and-ive-released-them-on-github/786094/4

I found that

[snow forge]

https://github.com/JimmyCushnie/Noisy-Nodes

Just found this that I think has promise. 🙂

lol. There's timing.

[stiff zenith]

We found the same thing lol

Tell me if you manage to get it to work!

[snow forge]

Yeah will do. Calling it a day for now, but will try it out tomorrow 🙂

[stiff zenith]

Ok

[true meteor]

Thanks a lot

[grizzled bolt]

!collab

[echo moat]

:loudspeaker: Collaborating and Job Posting

We do not accept job or collab posts on Discord.
Please, use Discussions to promote yourself as job-seeking, advertise commercial job offers, or look for non-commercial projects to participate in:
• Collaboration & Jobs

[normal rose]

Hello everyone somebody could help me? with line renderer component: i need make corners like this with line renderer

[viscid knoll]

To draw a square you’ll need 4lines

[young ice]

Hello! Is it possible to sample a value in a shadergraph from c# code?
I would love to be able to pass in a vector2 and get an output from 0-1 on this node.

Currently it's doing displacement to a plane and a raycast to get the height of the geometry doesn't work
(Heading to a meeting, back in 30, many thanks for any help)

[viscid knoll]

What are you trying to achieve? Is it about getting the modified physics data (transform height) from the raycast using the already displaced height of this terrain/plane?

[grizzled bolt]

I cannot tell from the wording if you want to move data from shader to script or from script to shader
Assuming from shader to script, it's not really practical
Commonly you'd rather run the same noise algorithms in a script just for that one point

[young ice]

Cheers guys,
@viscid knoll, kind of yes, I have rolling hills and I want to sample the height each frame at specific points. Traditional raycast doesn't work as it only hits a collider and I'd need to apply the mesh displacement to a MeshCollider.
I've assumed that the cheapest way to do this is to read from this point in the shader graph and then use that as an offset from the planes transfrom.position.y.

@grizzled bolt I was thinking shader to script, but if there's a way I can generate those combined gradient noise nodes in code and pass them into the shader, I'll have all the data I need to sample height in code.

The hills pan in a certain direction based on a UV offset passed into those Gradient Noise nodes, just not shown in the pic

[grizzled bolt]

It's not very practical or performant to get data from a shader graph to the cpu
Compute shaders are better at it but as I understand there's still a lot of situation where you'd rather have the scripts run their own noise generation anyway to prevent the hurdle of having to pass data between gpu and cpu

And it seems in your case if you do the deformation in a script, you might not to do it in a shader at all

[young ice]

That's an option yeah, I'd imagine that would be very performance heavy too though? Can't benefit from the power and parallelism of the gpu?

[grizzled bolt]

Documentation has the source code for all SG nodes, though just one gradient noise node is relatively expensive even on the GPU

[viscid knoll]

you have to sample the texture data from your raycast hit (1 point), then abstract the offset that this 0-1 value represent on the y axis to add it to the raycast hit result, and get the final "computed" height on this point

[grizzled bolt]

You could not, but the CPU is capable of parallelism as well which would be useful in this kind of task

[young ice]

Yeah I thought this to, I may be wrong but my understanding is there is no "texture" here, it's all proc-gen'd data fed into vertex displacement. I can't do any kind of material.texture.GetColourAtPoint(vec2)

[grizzled bolt]

Indeed, the GPU would have to write the texture first and then the CPU would have to spend the effort reading it

I do not know if that's less or more efficient than doing it on the CPU to begin with
But it also depends on what kind of noise algorithm you use and if you utilize multithreading

Efficient terrain generation algorithms are kind of a big topic so there should be many options more efficient than gradient noise

[young ice]

Yeah
My current working solution is to have a camera with a fov of 0.01. In front of that camera is a tiny quad with an unlit shader that draws Linear01 depth buffer data. That camera renders to a 1x1 render texture and I sample 1 pixel. I get a value between 0-1 and I multiply by the cameras far clip plane. That gets me an accurate distance, but it's munches ~20-30 fps, and that's just one, I need ~10

[kind juniper]

I'd still think that a compute shader would beat CPU by a big margin.
Though, until you test and profile it's just an assumption.

Generate your height map in a compute shader then use it both in the shader for rendering and the CPU side as you wish

[low lichen]

If you can reimplement the noise algorithm exactly on the CPU, then you can displace the mesh with the shader and then when you want to sample individual points on the CPU, just calculate the noise for just those points, no need to move everything over.

[grizzled bolt]

It also depends on how many points you need to sample
Generating the noise for 10 points in script is a different beast from doing it to every vertex

[young ice]

I read up on using a compute shader to do a raycast, but it required exporting the shadergraph to code and modifying that code. (The contents of that shader code was ~5000 lines and repeated itself 9 times, I'm not overly familiar with written shaders and didn't go further in this direction)

Interesting ideas, I'm not clued up on how to write that in a compute shader but I can give it a go 🙂

[grizzled bolt]

To use compute shaders you'll need to know how to write shaders and then how to write compute shaders as well

[low lichen]

I mistakenly said compute shader in my comment because I thought that was what you were using, but my suggestion also applies if you're doing the displacement in a Shader Graph shader.

[young ice]

Are you thinking a hybrid solution, the shader does the displacement and the cpu, when sampling the height, runs the equivalent noise generation and then samples the value?

[grizzled bolt]

Yes

[young ice]

I like it, I could likely even do some optimisations by generating the noise for the current region, then instead of generating it every frame as it pans, cache it and sample the noise with the offset applied to the sampling point

[young ice]

https://github.com/Unity-Technologies/ShaderGraph/tree/master/com.unity.shadergraph/ShaderGraphLibrary
Am I in the right place for getting the source code to the Gradient Noise node?

[grizzled bolt]

https://docs.unity3d.com/Packages/com.unity.shadergraph@17.0/manual/Gradient-Noise-Node.html you can just look them up in the docs for the calculation

[young ice]

Awesome, thanks!

[manic locust]

Hi. i need some help/guidance on a shader i need to make for a character. I have 3 skinned meshes. the third one is a blend of the 2 previous ones. Cause in the first object, parts have unwanted deformations. so i blend the proper deformations from object2. and the 3rd object has the final correct shape that benefits from physics sim and proper original skeletal anim.
so i explored many ways, compute (cant find how to update 3rd mesh's graphicsbuffer), manually set vertices (slow as hell).
I ended up using Vertex Shader in a shader file. in order to have the full features of standard shader, i started extracting bits of code from the downloadable builtin shaders. and ended up with a working test shader.

now that my vertex blending code works fine, i want to make the shader look good, so my question is do i have to modify all the passes and put my vertex logic in every single pass. Looks like there is deferred and forward stuff going on. i really have no idea what is what, its already a miracle if i managed to get to this point.

i made a post on the forum : https://discussions.unity.com/t/blend-2-skinnedmesh-vertices-positions-normals-in-3rd-skinnedmeshs-custom-vertex-shader-based-on-builtin-standard-shader/1510832

with the shader code, and the script i use to feed the meshes data.

i will apreciate if you can tell me how i can make all the shader work with that vertex blending code.

looks like i have to go through each pass and adapt.

[cosmic prairie]

camera has a lot of overhead, you could write a scriptablerenderpass in which you set up orthographic projection with cmd.setviewmatrix and cmd.setprojectionmatrix, set up the target texture with cmd.settarget and call cmd.drawmesh to draw the heightmapped thingy. then you can read back the data, or even better is to batch together a bunch of these into a 1*x sized texture that you will read back afterwards in one go

you could maybe even write the depth into a buffer too, that may be faster to read back but I'm not sure

use RenderDoc for debugging

[young ice]

@cosmic prairie interesting, sounds like it should optimise a lot. In my setup I have a lower poly version of the mesh which is displaced, no lighting, isolated via layermasks etc to try and optimise as much as possible

Shame we cant raycast polys without a collider...
Although I remember doing this in an old project, I had CAD data with ridiculous polycount, I used a grid of raycasts to pick up points on the mesh. There was no collider involved and yet the raycast hit the geometry. 🤔

[vague ginkgo]

I take it there's not an easy way to apply a vertex shader to everything in the scene with like a renderer feature? I was hoping to skew the whole world without applying it to every material

[low lichen]

No, that's not possible.

[red flint]

I've followed the shadow example in https://docs.unity3d.com/Packages/com.unity.render-pipelines.universal@16.0/manual/use-built-in-shader-methods-shadows.html
In the end I'm left with the shadowAmount value. How do I use this in conjunction with my existing fragment shader? I'm trying to match the color with the Lit shader but I'm not quite getting there

[flint yew]

This is usually the approach taken for ocean simulation

The GPU generates the height map for the entire ocean, and the CPU independently recomputes specific points to make objects float and whatnot

[young ice]

Ah of course, for simulating floating bouys on ocean surface!

[flint yew]

For instance

unless you have an obscene number of objects, its better than sending data from the GPU, although that can be done asynchronously, with compute shader readback

but direct recomputation is much easier

and intuitive

[prime plume]

In a script i am running lots of compute shader kernels. Each kernel modifies certain structured buffers for the next kernel to use, but this means that each compute shader is dependant on the last one to finish. How do i make sure the last one finished and do i even need to?

[flint yew]

i’m pretty sure compute shaders dont execute in parallell

[prime plume]

Hmm, weird. My issue might come from somewhere else then.

Basically when i started with unity i wanted to make a fluid simulation. Ended up making is 3d so i wanted a raymarching shader to visualize it, but it seems the data thats being sent to the shader is way different than what i get when i use GetData on the cpu and just visualize my voxel field with cubes.

[flint yew]

could be lots of things, i dont know

[prime plume]

Fair. imma keep on debugging, thanks for the help!

[prime plume]

I did a bit of research and i think that compute shaders so run asynchronously from eachother, but dont wait for eachother to finish (which makes sense), so i need to either pause the mainthread and wait for the compute shader to finish or i need to combine every kernel into one

[grand jolt]

I know Unity's shadergraph by default has limitations regarding the types of blending methods, is there a way to get an overlay blending with it?

Or even through normal shader programming

ive gotten some very close results but they can have the side effect of really fading out whats behind it

[flint yew]

really? mind sharing?

What if you use a command buffer, are they still run asynchronously?

[prime plume]

from what i've gathered no, it does keep its order in the buffer, but it doesnt account for synchronization so compute shader execution might still overlap with one another. i eventually just opted to use fences, but this greatly reduced my performance. idrk what else to do

[floral musk]

(URP) I am having an issue with transparency in build vs editor. The game I am working on loads a lot of assets from bundles in the cloud. One of these assets is a window. It is transparent in the editor when loaded (this loads from the cloud as well, the assets are not in the project at all). In the build it is not transparent.

The obvious solution is that its not including that shader variant. However I have both done a capture from the editor into a ShaderVariantCollection and added that collection to the list, but I have also added the shader to the always include list so it will build all variants. IT STILL IS NOT TRANSPARENT IN A BUILD. I'm not sure what to check next.

Using URP and the shader is custom built from shader graph

[low lichen]

There shouldn't be any specific keyword needed to make it transparent. Have you tried making a development build and connecting the Frame Debugger to it to see the keywords and blend state used for the draw call?

[floral musk]

no, thanks. That gives me a good avenue to continue testing with

[vague pilot]

Hey folks, im currently trying to implement a screen space reflection shader and i really cant figure out what i've done wrong at this point that this is the output i get (which changes with the camera pos) heres the code: https://pastebin.com/DGe1TPbw

[neat orchid]

for some reason, I can't find it online...

how can I add the tint feature to base maps via shadergraph

like with standard unity shaders

[tacit parcel]

the concept should be the same, multiply the base map color with a tint color parameter

[shadow kraken]

I usually make sure to make an extra lerp to factor in the alpha of the tint colour as well

[proud hemlock]

Is there a way to disable filtering (set to Point) on a Simple Noise? I want to get pixelated noise, not smooth like this (I am using URP, if that does make any difference)

[dim yoke]

No, the noises are generated procedurally without any textures being involved so there's no filtering being applied either. If you need pixelated effect, you need to round the incoming UV coordinates

[regal stag]

You can do UV node -> Posterize -> UV port on Simple Noise
Though using a pixellated noise texture might be cheaper

[proud hemlock]

oh, thank you very much! ❤️

[stoic flint]

guys is there a way through a shader, we make somehow the edges of snow mesh blend with the plane of the ground to create smooth transition or is it not possible through shaders alone ?

[broken valve]

is it possible to use shader graph functions in normal shaders if yes how

[regal stag]

Most of the node library docs have generated code snippets you could copy
https://docs.unity3d.com/Packages/com.unity.shadergraph@17.0/manual/Node-Library.html

[broken valve]

ye but where do i import SHADERGRAPH_SAMPLE_SCENE_COLOR from?

[regal stag]

The result of that macro depends on the pipeline. In URP it would end up using the DeclareOpaqueTexture.hlsl include

In short, sampling _CameraOpaqueTexture global texture reference

[broken valve]

ok thank you

[tired verge]

Guys, can shaderLab code be grabbed during runtime in the format as it is in unity?

[kind juniper]

Unlikely. It's all precompiled to regular hlsl(or other shader code)+ some data structures on the CPU side and then compiled to binary.

[vague pilot]

hey, i'm trying to make a screen space reflection shader and i think i'm missing the last step to make it work so i do not really understand how to check if the ray that's being sent towards the reflectionDirection is intersecting with geometry i would guess that you would check if the depth at the rayUV point is less than the distance the ray has traveled from the camera (so the "new depth") but how would you check this would you convert the rayPosWS into rayPosClipSpace and check if the rayPosClipSpace is further away? here's the code: https://pastebin.com/R7QuhBuA

[tacit parcel]

there was a thread about this in unity forum, but I forgot where, basically you need to get the ground height (worldpos.z) and the snow height (again, worldpos.z)
now the tricky one is you can't just get the ground mesh position from snow material.
But I think you can get the depth map using render texture with top down camera (by rendering depth map, excluding the snow using layer), then use the render texture as the ground height.
It wont be easy, but it can be done (I think it already is)

[tacit parcel]

Found it
https://discussions.unity.com/t/solved-zelda-botw-terrain-intersection-blending/760769/4

[stoic flint]

i will look into it

[regal stag]

add {} at the end of the first property (after "=white")
Not sure why textures need this, but they do

Yeah it's a bit weird, only breaks with other properties

[untold forge]

Hello! Apologies if this isn’t the right place.

Yesterday I was looking at Scanner Sombre and wanted to try and implement the Lidar Scanner myself. I did my own research and found that most people or tutorials online use VFX Graph which is perfect for that, but i want to just do it in shader code if possible.

I have the idea of creating a post process that renders the depth of what i see in a flashlight “cone”. Storing that in a render texture and passing that on to a compute shader where i read that RT with the depth info and calculating the world pos of that pixel (Obviously later on id have some noise to choose the pixel and not do every single one…).

Once thats figured out i use Graphics.DrawMeshInstancedProcedural to draw a mesh of a quad at its position in a C# script.

I know im over complicating myself but I just want to do it in shader code. Before I start i’d want to hear if this is a good idea or im going in the wrong direction.

Any advice is helpful. Thanks!

[tired verge]

Is this that precompiled data what we can see in Renderdoc as well when we check the shader source option?

[kind juniper]

Depends. You can see a lot of things in Renderdoc.

Shader source yes, that's usually in the language that the graphics api support

[flint yew]

Yeah, sounds fine.

What kind of meshes will u be rendering?

a bunch of tiny quads representing the points?

[untold forge]

Pretty much yeah

[flint yew]

It might be tricky to do the "scanning while moving" part

if youre just doing snapshots whenever the lidar is activated, thats pretty trivial

[untold forge]

I was just gonna do snapshots, i did think about the moving thing being a little bit different but we’ll see when i get there and if i want to do that

[flint yew]

I see the lidar in scanner sombre just shoots out rays instantaneously in random directions

so that seems easy enough

[untold forge]

Yeah i did think oh i can do raycasts or some form or raytracing but i just want to see if i can do it all in shader code except for the computebuffer part that will be in c#

[flint yew]

Render the depth, generate random pixel locations, dispatch a compute shader to calculate all the hit positions, generate quads.

[untold forge]

Alright well, i’ll give it a try! Thanks for the input, just wanted to make sure i wasnt going in the wrong direction

[flint yew]

Yeah, I wouldn't use ray tracing for this, just stick with the depth buffer

Though ray casting would be fine if youre not shooting obscene amounts of rays each frame

good luck

[dusky arrow]

hello Im trying to sample the skybox in shader like the code in the screenshot. But it is not correct in the editor, and when I check the framedebugger I see a grey result for the skybox
This is using a Custom RP based on this tutorial series by CatlikeCoding

[tame socket]

Okay, does anybody know why this operation isn't just drawing the main screen? I assumed that blitting back and forth would be identical to perform no blit at all, but for some reason it draws an empty scene as opposed to when nothing is performed and it draws the objects in the scene properly.

To illustrate, empty scene is with the code present. Scene with cube in the middle is with code commented out.

[warm moss]

Have you tried injecting your own command buffer into the render loop instead of using the Graphics class?

[tame socket]

I have not. I was just trying to debug some strange behavior in my script by checking the texture contents with blit when I found this oddity

[warm moss]

I'm just not sure the camera target is bound to the after post process texture outside the render loop.

[tame socket]

I tried printing out Camera.main.targetTexture, and it doesn't come back as null

Maybe that's the problem

The weirdest part to me is I have a similar piece of code in a different project that works just fine. This is all in OnRenderObject by the way

[tame socket]

I'm actually losing my mind

This makes no sense

[dusty creek]

random thought came into my mind and got me wondering, when are compute shaders a usecase? they intrigue me but I have no clue when they should be used

[tame socket]

For me, I'm using it for parallelized fluid simulation. While the rendering can be done through shaders, the actual math behind each particle's simulated forces can't be (as far as I know). ...I'm also using it for the rendering because figuring out Unity instancing in shaders is a terrible labyrinth that I don't want to figure out. But now I'm in this bind.

So basically, parallelized physics calculations I guess

[grizzled bolt]

When you want to utilize GPU cores in calculations outside of shaders
Simulations and generators are common uses, but the applications are endless

[dusty creek]

so like pathfinding?

or fluid sims

stuff that does a lot of math for many elements?

[dusty creek]

not sure what you mean by using it for rendering

[tame socket]

Yes

[tame socket]

I'm passing the RenderTexture texture to my compute shader to perform pixel-wise operations manually. There are a lot of ways to render fluids, and one of them is through screenspace operations.

So instead of having a regular fragment/vertex/surface shader, I'm manually calculaitng the RGB at each pixel on the screen based on input RenderTextures and bunch of other junk

[dusty creek]

oooh that sounds interesting

I'm guessing this fluid sim is for water? so ur doing like reflections and refractions?

[tame socket]

It can approximate fluid-like behavior, depending on the parameters, but yes I'm trying to simulate water. While reflecting and refracting is a part of it, a major portion is actually performing smoothing operations on a depth image of the water particles to approximate surface of a fluid. There's a million academic papers out there on this stuff with tons of differnet implementation methods, but I'm just trying to get this thing done rather than optimal

There's also like 3 major methods for fluid simulation where each have tons of depth in their own right.

The one I'm using is called "Smoothed Particle Hydrodynamics", but that's enough procrastinating. I need to figure out how to fix this thing

[dusty creek]

u got this

[kind juniper]

Blit just performs a draw call with texture A being passed as a main texture and texture B being the render target. There's no point in using it with either one of them being null.

[tame socket]

Null corresponds to the screen RenderTexture is what I have been lead to believe

Is that incorrect?

[kind juniper]

I can't see anything like that in the docs. It seems you can blit back to the screen if you set both to null.

Ah, no. If source is not null, but target is null

[tame socket]

This also

It's been a right pain in the posterior having to deal with the weird null exceptions because of it too :/

[kind juniper]

Yes, that's what I said. If the destination is null, but the source is not.

[tame socket]

Either way, it didn't change the outcome even when I swapped to using Camera.main.activeTexture/Camera.main.targetTexture.

[kind juniper]

I don't think that would work as it sounds like you're rendering the render target to itself.

I'm not sure what unity does under the hood, but in most graphics APIs a texture can only be either in read/write state or render target state at the same time.

I'm not really sure what you're trying to achieve anyway

[tame socket]

So I wouldn't be able to pass the active renderTexture to my compute shader?

[kind juniper]

If you just need to have the screen pixels in a compute shader, just set one of the cameras to render to a render texture and bind it to your compute shader. I don't see why you'd need to use blit.

[tame socket]

I was hoping to avoid adding a third camera as I'm already encountering issues with the secondary one I have right now

[kind juniper]

What does the second camera do?

[tame socket]

It renders only the particles for my fluid simulation, so that I can use the depth image for post-processing.

[kind juniper]

And what kind of issues are you talking about?

[tame socket]

Ah, apologies. I misunderstood. It's unfortunately not something I believe this discord would be equipped to solve. I'm attempting to develop a plugin for a third party vtubing software, and currently when I put the plugin into the software and kick everything off, my secondary camera seems to somehow kind-of commandeer the vtubing software's main camera status and erases all the GUI components of it.

It's one of the many things I was attempting to solve tonight, but it has yet to bear fruit.

[kind juniper]

Does the second camera render to a render texture?

Or the screen?

[tame socket]

Render texture

[kind juniper]

Hmmm... In this case it's weird that it has any effect on ui or anything else really

[tame socket]

Though I suspect there is some funky stuff going on with the copyFrom function I use to mimic the mian camera. Was going to test a bunch of things but many things broke and I tried catching 15 rabbits at once

[uncut gyro]

Hey, I have a basic tilemap which looks like this.
I want to make the transition between colors smoother with some noise applied on the transition. How can I approach this?

[grizzled bolt]

Shaders do not have information about what's around each pixel, so generally you cannot do blends like this
You'd instead probably want to have a rule tile that represents a transition gradient between two textures and use that to blend textures that the shader already knows about

[uncut gyro]

alright, thanks. I will see what I can do

[gilded nest]

How can we measure the performance of a custom shader? I have 2 shaders that achieve similar results, I would like to see which one is more performant.

[low lichen]

If you want to measure specifically the fragment shader part of it, the simplest way is to fill the screen with that shader, by placing a quad/plane in front of the camera, and then swapping it back and forth while measure GPU time in the profiler.

[pseudo kite]

Hey I'm new to unity and following Gabriel's tutorial on lighting shader graph but for some reason it isn't animated here is a comparison to show what I mean:

I added the time value and even tried setting the distortion speed higher but it doesn't work
any suggestions?

[pulsar oar]

How i can make my own hdri shader

I want to make the visor have a fake reflectivity, like a env ball, but i want it to be black since it feels weird when its in another color

A cube map shader i mean

[pulsar oar]

Nvm i figured it myself

[pulsar oar]

Theres any shader that give a cool glass effect? Because the ones i tested or give a plastic look or loads a random sky box, that idk how to change

I will try using mtoon vrm to test it

[warm moss]

You could render tile ids into a screenspace id render texture, then generate 1 pixel width outlines with a sobel filter, then generate an SDF texture with JFA. That gives you the distance to the edge. You can then sample that distance from the tile shader, normalize it and use it as a noise strength value. You need to sample the distance texture from your tile shader and multiply the noise value by (1 - normalizedDistanceToEdge). You can then output the resulting value as the alpha or shift uvs. Note that you have to keep the base tile layer without the noise and put noisy tiles on top or else you'll see holes in the world.

I hope you know graphics programming, otherwise it's easier to just draw some extra diagonal tiles and apply a global worldspace noise to uvs of all tiles with a shader 🙂

[pseudo kite]

Hey sorry for bothering you but mind if you can help me out?

[uncut gyro]

Nope, I know almost nothing about graphics programming. But its not a problem :) New challange

[warm moss]

I'd go with the latter of my suggestions tbh, way less r&d.

[jagged lodge]

Any reasonable way anyone knows I can add a component to my render pipeline to discard front faces behind back faces? I would like to discard the red highlighted area but show the green outlined area, in this example.
I can think of a couple ways to write this in a shader for a material that attaches to the object, but I'm not sure about a way that applies to all objects regardless of the material on it.

I can imagine maybe inverting the mesh in a render pass and drawing the background color over the new front faces, but I would like to be able to overlay this over another camera and I don't believe this would work

Some way to get backfaces to write to the depth buffer globally maybe?

[quartz sail]

I have this shader graph, i made this toon shader, but im kind of new so i dont know how to add a normal map input to this?

[regal stag]

Sample the normal map with Sample Texture 2D node set to Normal mode. Use Transform node to convert from Tangent to World space, replace Normal Vector node in the Cel Shading group with that result

[quartz sail]

thank you!

[waxen merlin]

Sanity check - If you need to do multiple passes with the same compute shader, can you use the same instance and just change the inputs between Dispatch() calls? I assume that's true... that the Dispatch call sends off that particular invocation of it with all the settings. Also, given the passes read/write from the same textures, I assume Unity makes that safe, i.e. 2nd pass doesn't start until 1st is complete?

[hollow wolf]

is there anyway to add #include example.hlsl in the shadergraph? i know how to add #define by using keyword -> boolean/enum, but i dont know how to add custom #include

[regal stag]

Can use Custom Function node (File mode)

[hollow wolf]

ohhh, looks like i overcomplicated it, thanks for the insight 💡

[honest yacht]

Hey, I have a question, maybe someone will know the answer. I followed a tutorial to create a shader graph that creates a puddle like the one in the screenshot. However, I just noticed that the floor texture is scaling along with the water texture, and they don't match at all. Rescaling the puddle causes the issue where water is almost invisible. Is there a way to remove the floor texture's border and just keep the puddle?

I tried adding a regular transparent texture, but the area around the puddle turned black

[drowsy shuttle]

Hey thats a long thing to tell and in that case i would say get the sample shaders from unity it includes puddles,ice,decal and more.
How to get:
Go to packages then shader graphs then navigate to sample and scroll down and then import production ready shaders.
Note: this is available in version 2022.3.35f1 or higher.

[quartz sail]

this might be a dumb question, but am i able to animate shaders graphs to happen on an event, or do i need to code the shader and do it from there? for example, a disolve shader happening on death or teleport or something,,, or would this be better done with a vfx

[civic lantern]

The simplest way is to have an exposed float parameter in the shader, then control that value from C#

material.SetFloat

[quartz sail]

i see isee okay thank you

[civic lantern]

VFX graph can be good for this kinda stuff too, just depends on how complex you need it

Like if you need it to dissolve into particles or something

[quartz sail]

yeah i was just thinking about that, i think it may just be better done with a vfx,, im just better at shader lol

[rare wren]

Has anyone gotten soft shadows (incl quality tiers) working in Unity 6?
I got a good system for 2022LTS but it completely breaks in 6, while the generated keywords are the same as the Simple Lit shader

[quartz sail]

why is any of this happening, the normal map isnt doing anything crazy like this

[kind juniper]

Is it fixed if you remove the normal map?

[quartz sail]

yes it is, but the map looks fine i think, i redone it multiple time so i really dont knwo whats wrong with it

[regal stag]

The normal map icon is very small so hard to see but the colours don't really look as I'd expect. To clarify, Unity uses tangent-space normal map (the blue/purple looking ones) not object space.
Make sure you import the texture as the Normal Map type too

[quartz sail]

that must be the issue, am i not able to use object space maps at all?

[regal stag]

I think you can with a custom shader but I'd recommend just baking tangent space ones if you can

[quartz sail]

alright thank you

[kind juniper]

Just to clarify, you do import the texture as a normal map, right?

[quartz sail]

I rebaked it as the tangent space and it looks fine, I normally just drag it in from my files and then after putting it in the normal map spot in the material inspector hit the “fix it” button from the “this isn’t considered a normal map” 💀 is that bad

[kind juniper]

I guess that does the same thing. In the past, unity wouldn't show you that warning, so you had to make sure yourself.

But it's always good to have a look at the asset import settings before using it. There are a lot of settings you might want to tweak.

[quartz sail]

Alright thank you!

[lapis thorn]

Are there any shaders that have specular setups that are very similar to Unreal Engine 3 and alot of other old engines' specular shaders? Unity's own specular shader looks extremely odd, the specular map shows all over the model and is way more visible than it should be.

[kind juniper]

The only difference I see is brightness 🤔

[lapis thorn]

Well here's the specular map, in UE3 the specular map is only supposed to show a little bit even if the scene is bright or dim. In Unity, the specular map shows 100% and in general makes my material look ugly

Not sure if that entirely makes sense with the way I explained it

[kind juniper]

This doesn't look like a specular map to me. Looks like albedo/color texture.

[lapis thorn]

This is the color map

[kind juniper]

A specular map just decides how reflective the pixel is where it is sampled.

At least in unity.
If it means something more in unreal, you'll need to find the alternative in unity.

I guess it also depends on what kind of shader you're using.

[tacit parcel]

Both unreal and unity has switched to pbr, where you dont manually define specularity anymore.
But you can use the legacy shader family which has bumped specular shader.. iirc, the specularity is on alpha channel

[lapis thorn]

Alrighty, thanks for the help anyway

[vestal hinge]

Hi all, I'm writing a custom shader which is called through the Graphics.RenderPrimatives function, in this I am trying to implement occlusion culling. That said I seem to be running in to a bit of an error https://youtu.be/G9L6SRxFJls (shown in this video). The code I am using to do this is below, to explain I am using the _CameraDepthTexture sampler from UnityCG.cginc and then comparing it to the active depth of the fragment being sampled. This seems to potentially work fine when the camera is still but moving it causes huge artefacts.

const float2 screenSpaceUV = i.screenSpace.xy / i.screenSpace.w;
const float sampledDepth = SAMPLE_DEPTH_TEXTURE(_CameraDepthTexture, screenSpaceUV);
const float depthBias = 0.1;
if (i.depth + depthBias < sampledDepth)
{
    discard;
}

return i.color;

I am definatly doing something wrong here, so if anyone knows a potential cause for this I would really appriciate the help!

[flint wasp]

This might be because your depth texture is being copied after your shaders fragment pass is processed. The active depth exists on a depth buffer, to make sure you have a thread safe copy to read it’s copied at some point to _CameraDepthTexture. In URP you can set when this is copied in your renderer asset. Idk about built-in though.

If you want this shader to cull geometry also using this shader behind it you need to define a custom depth pass so you can first write the minimum depth to the buffer then cull the geometry behind it in the lighting pass

When you stay still what I assume is happening is that it’s reading the depth from the previous frame which is why it looks ok. I’m not sure though because usually built-in depth texture is cleared between frames

[gloomy wasp]

In the built in render pipeline, the internal deferred shader uses a PBS lighting model. Is it possible for me to make my simple toon shader compatible with Unity's deferred rendering even though its not a PBS model? Or would I have to swap out the internal shader with my own solution?

[low lichen]

Toon shading isn't really possible with deferred rendering, because of how each light is added additively on top of each other, so they can't be posterized together.

Unless you render the lighting to an offscreen buffer and do the posterizing on that, I guess.

But I don't think built-in's deferred rendering is customizable to that degree.

[flint yew]

what about a post process?

is that a thing

[cosmic prairie]

if you use URP or HDRP it is possible to edit the package to make it do the toon effect

built-in is outdated, don't use built-in in 2024 unless you know what you are doing and why you are doing it

[vestal hinge]

Ah thank you! I'll look into this, I am using URP so I'll see if I can find what needs to be changed in their!

(Sorry also for the late reply, I ended up falling asleep xD) (It was quite late where I am when I asked the question)

Changing the value to force prepass seems to have fixed the movement issue! I have noticed though that my eyeballing on the two depth values wasn't actually correct. I've taken the differences between the two depths which should be black (if they where the same, but its clear that this isn't the case.) I'll look into this now, thank you for your help!

Okay take it back, the issue with the difference was that the new output of the prepass gives a completely black value. so the difference is always just the depth value of the actual pixel. Moving it back to the AfterOpaques, gives me back the black transparency (which is correct, but this then also has the issue of bringing back the movement issues again).

I am currently calculating the depth using this code:

const float dotValue = dot(wNormal, normalize(_CameraWorldPosition.xyz - worldPosition.xyz));
float4 clipPosition = UnityWorldToClipPos(worldPosition);
o.depth = clipPosition.z / clipPosition.w;

Potentially the distance to the camera is what is failing. I am passing in the position of the camera right now in late update, is their a chance that this could be the cause? And if so, where in the excecution cycle should the code sit for uploading data from RenderPrimatives?

[vestal hinge]

I did a few more tests with this, and it seems the the depth buffer is slightly behind of the cameras real position

This was taken my just screen shooting the game as I span the camera around, its pretty clear with the opening of the "cave" that it is being mirrored to some degree.

So I think the assumption that the depth texture sample is somehow lagging behind on the sampling.

Right now the depth texture is being taken after trans parents this could be that the pass is using the last frames depth test

Sampling the prepass depth however just gives a black value (so zeri in everything)

[inner hare]

Do shaders only work on UI objects that are in world space?

[flint wasp]

@vestal hinge
Ok a few things: the reason a render texture is black could be for two reasons. 1 it’s copied at a later stage then when you are readin it or 2 there is nothing writing to it.

When you force a prepass for a normal shader, Unity would first look for a depth pass on your shader, and will use your default lighting pass if nothing is found(I think). This creates two commands to render the mesh(one for depth, another for color) at two seperate render stages. This becomes a lot more complicated with renderprimitives.

When you call render primitives the shader is not run immediately, but the command is queued in the pipeline at the specific render priority https://docs.unity3d.com/ScriptReference/RenderParams-rendererPriority.html you’ve set in your render params. What you need to guarantee is that by that time your shader is run the depth texture exists and is populated. You can check which priority maps to what pipeline stage.

If you use a depth prepass(which you probably must) for geometry also rendered using the shader, I would assume you also need another renderprimitives command with a lower priority equivalent to when the depth prepass is run with the material having set the depth pass to render; https://docs.unity3d.com/ScriptReference/Material.SetPass.html. You can probably do this or make another material just with the depth logic.

I’ve never personally tried this but this is what I assume you would need to do. There may also be a simpler way

[grizzled bolt]

There's a whole breakdown here: https://youtu.be/Z_-am00EXIc

[vestal hinge]

Thanks I did manage to solve this by adding the pass, you where right in saying that render primitives dont add to the depth buffer on force prepass without specifying an extra pass before hand. I was wondering however, is their a way to have the depth texture used in the shader remain the same between cameras? Specifcally I am trying to see what is being rendered using the editor camera - but since the depth buffer is different for this camera compared to the one that is rendering the game the editor camera renders things that should be obscured to the game camera.

This makes complete sense, but at the same time it does make debugging a bit of a pain. So I was wondering if their was a way to force a shader to always sample a specific cameras depth buffer instead of just _CameraDepthTexture being the actively rendering cameras

[grizzled bolt]

They are post processing effects as the video explains
The shaders on the materials don't look to be anything special, probably just HDRP lit
In fact the video proves it's HDRP Lit so why don't you watch it

You can posterize anything you wish

That should start becoming clear when you study what the difference of materials, post processing and renderer features are

[dapper heath]

heya, just a quick question about rendering related stuff- I'm using URP atm and I have a shader graph I'm using for my scene that handles outlines and etc., and right now I have it applied per-material. so everything has a copy of this shader and sets its parameters piecemeal. would it be better performance-wise to have a fullscreen shader apply the outlines and only have each material handle stuff like textures & normals?

[kind juniper]

To clarify: now you're using one shader that does both normal rendering and outlines? Or you have the outline shader rendering the outline only in addition to the normal rendering done by other shaders?

[dapper heath]

it's all one shader, just split up into diff. materials (one material for a house using a trim sheet, another material for the player, etc.), and i'm trying to figure out if it'd be better to convert the outlines and other fancy parts of this shader into a post processing effect

then just keep the standard texture stuff on the per object materials

better performance-wise, i mean

[kind juniper]

No one would be able to answer that really. You could test and profile different options to see the difference.

[flint wasp]

Unfortunately I don't know how to override which camera writes to the depth buffer. This probably is possible but probably also requires a lot of messing around with the pipeline. The question is, why would you want to even sample the depth buffer from the scene view? If you use the game camera's depth to cull geometry traced from the scene camera's view, the result will make no sense. If you want it to completely switch to the game camera's view, why are you even using the editor camera?

[vestal hinge]

Ots mainly so that I can visually debug how much culling is being applied on the game views camera but from the angle of the scene view.

Right now thw game view looks pretty good but from the scene view parts of the geometry is visible that shouldn't be. My first thought was that I'd messed up the occlusion culling. But I realised the depth texture is using the scene view (editors) camera to create the depth on that pass of the camera.

I know it's a niche use case and I won't use it in the actual game but debugging wise it will help alot with finding out what things need to be optimised and where.

[flint wasp]

Ok that's pretty weird. Uhh if Unity documentation is to be trusted at face value, RenderPrimitives' render params has a parameter for camera, which I guess might be what you're looking for.

There's two possibilities for what this means, either it means render from that camera's settings to the current render target--then it should do automatically what you want. If it means render using that camera's settings to a different render target then you probably can just find that texture and blit it to _CameraDepthTexture.

[vestal hinge]

Thank you, I'll look into seeing if this works! Yeah its a pretty weird use case but thanks for helping me with it xD

[flint wasp]

Np 👍

[fluid peak]

so, odd thing I ran into tonight... in 6000.0.11f1 .. in shadergraph the "posterize" node has no effect on voronoi noise in the scene. the previews inside shadergraph show it broken into steps perfectly fine, but when viewed in the scene its still 100% smooth noise, like the posterize is skipped completely.

[flint wasp]

Ohh I get what you’re saying now. You want to depth occlude the vertex by the depth from the game camera while the geometry is drawn to the scene camera’s screen. This may look weird because the not-culled geometry’s normals would face the game camera, so the scene camera might see backfaces

[vestal hinge]

For sure, I will look really dodgy but thats kind of expected xD
Ill send a photo once I get it working, ended up getting a bit side tracked with some generration side code. xD

[flint wasp]

Please do! I want to see what it looks like.

[grizzled bolt]

Did you save the shader graph?

[regal stag]

I've encountered this in older versions too. I don't know for sure why it occurs though, possibly something with the double loop that Voronoi code uses and the way the compiler decides to handle things.
I've found that using a non-integer value for the Steps or Multiply by 0.01 inbetween "fixes" it

[fluid peak]

yes it was saved and recompiled

@regal stag Thank you your suggestion of non integer values worked but its still strange 🙂, i guess just a quirk to work around

[supple kayak]

I need some help with a low poly water shader that I have created, there appears to be something weird happening with the waves on some places and I have no idea how to fix it, and since I know there are people smarter than me with shader I thought I would give it a shot 😅
I also have a noise subgraph and I added that to the list of pictures. Any help is greatly appreciated

[amber saffron]

I don't see anything in the vertex part that could cause "holes" or something like that.
Maybe it has to do with the flat shading, is it still visible if you don't write the normal output ?

[supple kayak]

you are correct, the problem is gone if I dont use the normal output, so I suppose its the way I create the low poly effect. But this is what I see most people do so im not really sure whats the exact issue

[amber saffron]

Hum, why is the transform node saying (Legacy v0) ?

[supple kayak]

Ah because, i found the shader graph for the flat shading from someones github and I suppose they created some time ago and Unity came with new Transform node, but I changed it to the default Transform node that unity uses now and with the same settings the problem is still there

[amber saffron]

Does it also happen if you only displace on the Y axis ?

[supple kayak]

um, do you mean if I take the Transform output, split it and only use G value?

[amber saffron]

No, in the vertex stage, don't add the noise directly to the position (this adds it to X, Y & Z), multiply it with [0,1,0] first

[supple kayak]

Oh you mean to multiply the position with [0,1,0] and then add the noise stuff?

[amber saffron]

No, multiply the noise with [0,1,0] (or, make a [0, noise, 0] vector), and add that to the vertex position, to have only a Y displacement.

[supple kayak]

oh im sorry, please bare with me i am not used to shaders lol
The image shows what I did, however the terrain has just dissapeared

[supple kayak]

aaaaaaaaaa see now i get ya hahaha

[amber saffron]

(you can also of course still mutliply with the water height)

[supple kayak]

Youu can still see the problem however 😔 , why am i so bad at shaders

[gloomy wasp]

Im learning about the built in pipelines deferred rendering and I wanted to make sure im understanding something correctly :D

Can my materials have individual shaders that do their own things as long as I add the lighting later? Like I could have a shader on a material that changes the albedo or something but the lighting is added later by the internal deferred shader?

[amber saffron]

Yes, that's what surface shaders are for.

[amber saffron]

I'm having some hard time to understand the source of the issue, the setup seems correct to me 🤷‍♂️

Could you maybe send a zip with the graph and subgraphs, as well as your mesh ?

[supple kayak]

Yeah sure! Give me a moment

@amber saffron here you go, and I also want to thank you for taking the time to help, it means a lot😅

[amber saffron]

Ha, found the issue, it's because your shader is transparent ^^

[supple kayak]

What? That was the problem haha

[amber saffron]

So, those darker polygons are in reality the back faces of your water, showing through the front faces, as transparent objects by default don't "self depth test" to put it simple.

[supple kayak]

Not at my computer right now so I cant test it lol but if it works for you

Aaaaaaaa that's what it is

Ah man thanks, so do I need to set it to opaque?

But I Kinda need the transparency since its water, and it usually have a bit of transparency. Do you maybe know a way around it?

[amber saffron]

Set the "Depth Write" to "ForceEnable"

So it is transparent and writes to depth for self intersection, like opaques

[supple kayak]

Aaaaah alright I will definitely try that once I comeback to my computer, but if it worked for you it should work for me aswell hahaha

Huge Huge thanks man, appreciate a ton, you saved me alot of tears hahaha

[spiral lagoon]

Hi, I need some help with a shader, the idea is to make a "2D Shadow Shader", it is just a regular shader but, if 2 pixels using the shader overlap, the latter one that overlaps gets skipped.

That creates the problem that you can see in the image, highlighted in red.

Problem: When a sprite is in between the 2 shadows, in terms of layer ordering, the pixels at the top shadow sprite don't get rendered.
That's because under the default sprite, there is a shadow sprite, which makes it skip the render. How could I fix it? Is there another approach?

The set up is the following:

• Shadow Sprite (Layer 0)
• Default Sprite (Layer 1)
• Shadow Sprite (Layer 2) --> Skips rendering, because at the bottom there is a shadow sprite, even though it is obscured by another default sprite pixel.

This is the current shader, with the sprite sandwich problem. :/

White Circles: Material with Default Sprite Shader -> Layer 2
Black Transparent Circles: Material with ShadowShader -> Layer 1 / 3

[gloomy wasp]

Great, thank you for your time!

[slate path]

What needs to be present in a (hdrp) shader so that it respects hierarchy render order when used on a UI image? I'm trying to understand why the shader im using ignores the hierarchy entirely (ie, an image that is above it in the hierarchy and should thus be maked by it is instead rendered on top of it)

[rose bronze]

Anyone got any experience making an interactive 2D grass/foliage shader in unity?
I have spent days googling and everything is 3D, which is not what I am after.

[grizzled bolt]

That's a bit vague, there's a lot of styles of 2D games with potentially different implementation
The same principles should generally apply to 2D that apply to 3D regarding vertex shaders specifically, as long as you are aware of how sprite renderers handle geometry generation

[inner hare]

Can you disable shader visibility in the inspector?

Like as if it were a Gizmo?

[grizzled bolt]

Shader visibility? There isn't really much else to see besides shaders

[inner hare]

My apologies, the refreshing of shaders like ones that depend on time for example

I see Always Refresh but it only stops the Scene refreshing, and it continues with mouse movement

[grizzled bolt]

I don't think the editor really knows what's different about a shader that uses time and one that does not
My solution might be to make a custom time variable which increments at runtime, or optionally also in editor when permitted to do so

[inner hare]

Ah okay I see. Thanks Spazi

[dense dagger]

how would i go about making a shader that displays an outline around an object in 2d

[tacit parcel]

by sampling neighbor pixels, check if the alpha is different with the current one, draw an 'outline' pixel (by setting color to outline color and alpha to 1)

[strange prairie]

I think shader graph has a UI preset.

[slate path]

Not using shadergraph. After some more testing it looks like it’s only an issue when using a screen space - camera canvas. Works fine for screen space overlay

[strange prairie]

IIRC screen space overlay is the only one that works properly in HDRP

[slate path]

Great, sounds like a catch-22 then. I need post processing (hence camera canvas) but the shaders only work properly in overlay

[amber saffron]

Re-do the shader in shadergraph with hdrp/canvas target ?
Or if you really want to code it, make a barebone shadergraph, save the generated code and edit it.

[strange prairie]

Yeah would just look at the generated code to see how it's doing it

[ruby nova]

hi

im havin issues w/ clipping n stuff

this cutoff thing is happening because i enabled ZWrite to fix another issue, and I have NO idea how to fix this...
im using a downloaded PSX style shader btw

[sullen mauve]

can anyone help me get rid of this artifacting? i don't know if this is a shader or a mesh issue

this is my vertex offset setup. i have no idea about how to do this so this might be very wrong

[hearty obsidian]

I'm trying to get 0-1 coordinates over the span of a 9 sliced image to map a texture over the surface.

Since I cant use regular uvs, I'm dividing the object space position by the mesh size. The mesh size I input manually, and I confirmed that it is the actual size of the mesh by sampling the verts from the canvas renderer attached to the image.

This should give me -0.5 to 0.5 coordinates, and then remapped should give me 0-1 coordinates like I want.

However it seems like vertically I have an artifact. Anyone see anything?

Seems like vertically the coordinates go beyond 1

the light reflection on the button is the texture I'm blending with the fake uvs

[hearty obsidian]

err, seems like object space position does not mean what I think it means for GUI objects

[viral pelican]

so im not into shaders and stuff like that and i'm experiencing a problem that i couldnt find solution for so i thought i will ask someone who has more experience than me. I have a retro effect shader that applays vertex jitter, and also i have a fog effect that comes with a shader and when i use it with the retro shader there is a wierd thing going on it seams to me that the fog isnt beaing renderd behind objects and i would like someone to help me fix that:

there is the fog shader:

and the fog script thats attached to the camera

i would love for someone to help me 🙏

[grand jolt]

I am trying to use a displacement map on a textmeshpro in order to create a glitch-like effect. Any recommendations on how I should go about this?

[tacit parcel]

try adding "addshadow" to the pragma in your vertex jitter shader

[zinc aspen]

@grand jolt

[viral pelican]

I'll try that

[grand jolt]

Hey, I need to ask a small question for Unity 2019.
I'm modding MPN (we have an official devkit project), and I'm trying to modify one of the toony-shaders to allow transparency, for the sake of a new type of asset.
Unfortuantely for me, I am not competent at shader logic, and keep getting either a unexpected '}' or unexpected 'end' error.

i dont really understand why closing the shader is considered unexpected.. could someone please help?

the rest of the shader is commented out right now to avoid issues

[amber saffron]

You need to add {} at the end of the main texture declaration (after "white", look at the other texture lines)

(or at least, I think that's the issue)

[grand jolt]

it's definitely a step to the right direction....
yet the unexpected closing {} on line 25 persists....

surely, if we gather enough monkeys we can type-writer this together

or one, very smart shader programmer who happens to have too much time on their hands

haha

[amber saffron]

Well, at that point I think that the shaderlab preprocessor is just loging an error because the subshader is empty 🤷‍♂️

[grand jolt]

i... yeah let me check into that. thank you 😅

[heady violet]

Anyone know anything about Compute Shaders? I cannot wrap my head around uint3 id : SV_DispatchThreadID because I don't know how the wave front distribution will assign it. I want to reference an index from an input buffer, then assign a property in a RW buffer of a custom struct. It isn't working, but the C# side looks fine and the HLSL shader also handles the input as expected, so I'm assuming that I am not using the id correctly in the compute shader and breaking it, not actually changing the buffer value or changing the wrong index.

{
    int Index;
    float Selected;
};

uint SelectInputCount;
StructuredBuffer<GpuSelectCommand> SelectInput;

[numthreads(64,1,1)]
void Select (uint3 id : SV_DispatchThreadID)
{
    if (id.x < SelectInputCount)
    {
        GpuSelectCommand command = SelectInput[id.x];
        Members[command.Index].Selected = command.Selected;
    }
}```

Any insight would be very helpful

[low lichen]

Can you show how you calculate the number of threads to dispatch from the C# side?

[heady violet]

just this MyShader.Dispatch(m_kernelSelectId, 1, 1, 1);

[low lichen]

The way numthreads and the Dispatch x, y, z thread count are combined is with multiplication. So, dispatching (1, 1, 1) threads, with [numthreads(64, 1, 1)] will dispatch (64*1, 1*1, 1*1) threads. id will range from 0 to 63. If SelectInput always has 64 elements, this is fine, but I think that's unlikely.

[heady violet]

So as long as I have less than 64 elements it should be working fine, right?

[low lichen]

I can't see where Members is defined, but yes I think it should work.

There's a possibility for a race condition if any of the GpuSelectCommands reference the same index.

[heady violet]

It's a controlled test so I'm sure that isn't the case. What would the function do if I had more than id.x in the input buffer? Does it start over at 0 ?

For example if I passed 50k commands into the buffer, how would I make sure those were distributed correctly and refer to the correct index in the Members buffer?

[low lichen]

You can just imagine the kernel function being called once for every id, from 0 to 63. It won't be able to read any commands beyond index 63, because the function is never called with that index.

If you think about numthreads as the batch/group size, you have to calculate how many groups you need to dispatch, by dividing the number of elements with the group size and rounding up.

This is usually done like this: Mathf.CeilToInt(elementCount / 64f)

[heady violet]

Ah that makes perfect sense. Didn't even realize that was an issue, thanks! Is there any way to debug whats happening in the compute shader? For example if I could log out what was in the buffer I could easily see if there is some issue with my inputs. (they look fine in C#, but who knows)..

[low lichen]

I recommend using RenderDoc. It's an open source graphics debugger that Unity can integrate with and has full shader debugging and resource inspection.

[heady violet]

I'll look into that, thanks!

[grand jolt]

did you mean to at me?

[strange basalt]

sadly it does not provide any for you, i did manage to do it anyways with a component that just adds a other UV set to the object that lie in the 0 to 1 space over the whole object then i just use UV 1 for 0 to 1 uvs and uv0 for 9 sliced uvs in teh shader. The reason object space is no good is because of the atlas. Can share the component i made for this if needed or explain generally what i did

[tame topaz]

They're just pestering people for attention, they've been doing it since they got here. They'll be kicked soon enough, so you can ignore it.

[grand jolt]

ok thanks

I have two options for doing this effect. One, I can use a render texture for just the text and a camera with it and then apply the shader to the text or I manipulate the text mesh manually with a script.

because otherwise im not sure how to get a good effect like this without the text behaving weird (letters turning all boxy and overlapping etc)

[strange basalt]

whats the TL;DR for what you are trying to acheive

just the glitchy effect on it?

[grand jolt]

the image i posted

yeah

i cant apply shaders to the ui toolkit so im trying to apply it a textmeshpro

[strange basalt]

yeah i would likly handle that with a render texture, then distorting its UVs when blitting

[grand jolt]

yeah im just wondering if thats overkill for what amounts to a death screen

or if i should handle it through script

[strange basalt]

it depends, pretty sure if you did it in the text shader it would just overlap and get weird

might be able to do it if converting to screenspace first before pushing things around

[grand jolt]

yeah its confusing because ive seen stuff like https://www.youtube.com/watch?v=FgWVW2PL1bQ

[strange basalt]

yeah moving stuff around in the shader is not that hard, the intro of that looks like it was using the vertex shader

and moving every vert around equally

so then no overlap issues happen

[grand jolt]

yeah i just need to find a way to apply that to a shader

[strange basalt]

oh weird they are doing it in code

but yeah doing it in shader, is really just modifying the vertex positions

[grand jolt]

i had a displacement map i was hoping to use but i dont think its practical given how little information i can find on really modifying textmeshpro shaders

[strange basalt]

its a large shader but would just start with a copy of the TMP_SDF.shader

most what waht you want will just be modifications in the VertShader of it

likly just modifying input.position before it does other work with it

[grand jolt]

ill probably make it simple until i can figure out a displacement map solution

ie: mapping red to horizontal and green to vertical

[strange basalt]

for displacement since its not height i would use a texture that is red modifies x green modifies y

[strange basalt]

would also convert it back to -1 to 1 space in shader so it can displace in both directions of each axis

[grand jolt]

im wondering if i need to split the map up too

because otherwise it would probably try to apply the whole thing to each letter

instead of to the overall text

[strange basalt]

depends on how you sample it

if using the UVs provided yes it will get weird

[grand jolt]

like if there is a way to directly apply an effect to the whole text that would be nice

[strange basalt]

but i would sample the texture using screen space coordniates

[grand jolt]

instead of the individual characters

[strange basalt]

sample the texture in screenspace would be the way

[grand jolt]

should i just target the font atlas or something

[strange basalt]

no talking sample your displacement texture in screenspace

would not change how tmp samples the atlas

Aside from adding offset

Based on the displacement

[grand jolt]

yeah but im wondering like how to actually apply it to the text i mean

unless the render texture route is the only way

i want to be able to affect the text for this but not have it like overly complicated and bloated through a setup like that

considering its not something thats always going to be in use

[strange basalt]

By useing your displacement to add offset to how tmp samples it's atlas

[grand jolt]

i need to figure out how to do that in a way that doesnt like end up grabbing other letters by accident

as i was facing that issue before

[strange basalt]

yeah that might still happen, what i would do is sample your displace/distort texture in screen space, bring its values into a -1 to 1 space then multiply it by a small number you can use to control the strength of the distortion

then when tmp samples its texture you can add this number to its UV's

[grand jolt]

im surprised there arent existing solutions already with how long textmeshpro has been a thing

[strange basalt]

there might be, but also not sure how someone would be a performant generalized solution

since hows its done depends on the wanted outcome, and if that sits within the limtations of a given approach

[grand jolt]

it wont be used constantly so there shouldnt be an issue with that

thats why i dont want the render texture solution

[strange basalt]

with my suggestion if the amount of distortion is not so great it pushes things outside of their polygons it should not be too bad

but it does have that limitation

[grand jolt]

yeah ill try to give it a shot

im still new with shaders

[strange basalt]

it gets much easier with experience

[grand jolt]

here was the issue i had earlier with my last setup lol

[mortal osprey]

What is the best ps1 style shader?

[tacit parcel]

why ps1 shader?

[mortal osprey]

https://github.com/Kodrin/URP-PSX I'm trying to use this but idk exactly how

[quartz otter]

Hello! I am trying to make alternate color palettes for my characters (for context, I am working on a 2D fighter). Could someone help me get started?

[tacit parcel]

I dont think thats how ps1 looks. Trust me, I'm an engineer I grew up with psx
ps1/psx looks are

• low poly, not those untextured flat shaded things, but really just low poly count model
• unfiltered low-res texture
• per vertex lighting
• affine projection
• 24 bit color mode. yeah, those posterized color stuff? that's not how ps1 look

most of them can be handled by assets alone, except for the affine projection stuff, which can be added by adding noperspective command to the uv declaration
https://danielilett.com/2021-11-06-tut5-21-ps1-affine-textures/

[mortal osprey]

Low poly + low texture res + post processing effects

[mortal osprey]

Oh but there’s more

[tacit parcel]

there is no postprocess in ps1, the console just dont support it (at least not natively/ootb)

[mortal osprey]

but we are trying to simulate the effect here

that's what people do

[tacit parcel]

what effect?
the wobbly texture is due to affine projection (modern games use perspective projection, which eliminates the wobbliness, ps1 just doesnt have the power to calculate it), some games try to reduce it by subdividing the polygons.
but if you want to simulate it using post process, then why not? there are many ways to simulate it

[mortal osprey]

in my case, I want to use the wobbly effect with low res textures and low poly
but to be perfect It need some blur and bloom to looks like a crt tv...the old tvs have this nice blur that merge the pixels and makes everything beautiful

oh and pixelated image...but I don't like the pixelation without blur

the blur is what makes everything harmonic and satisfying to look

[tacit parcel]

I grew up with psx and crt tv, and honestly, I never think it's beautiful 🤔

[mortal osprey]

I grew up with mega drive before psx...try to play it on an emulator today

It looks awful without the crt filter

I hate pixelated image

[tacit parcel]

I grew up with atari and nes...
and latest emulators usually have smoothing filters like eagle, lanzcos, xbrz etc for those who doesnt like pixelated image, like me...

And for perfect crt tv simulation, you might need some fish eye shader to emulates the curved crt monitor, then emulating the separated and offset rgb channels

iirc, some people already made those crt emulations, but as I dont really into those stuff, I didnt bookmark them

[mortal osprey]

the link I send here is a shader that do everything that you are talking

[tacit parcel]

and?

[mortal osprey]

and what?

[tacit parcel]

I never said I need those kind of shader

I just pointed out that thats not how ps1 games look

[mortal osprey]

of course it is

[tacit parcel]

well, if you say so... I have better things to do

[lusty verge]

isn't the wobbliness because the floating point inaccuracy on psx?

"One of the main reasons for the “wobbly” or “jittery” graphics often seen in PS1 games is the lack of floating-point precision. The PS1’s graphics processor did not have a Floating Point Unit (FPU), which meant it relied on fixed-point arithmetic instead of floating-point calculations12.

Fixed-point arithmetic is less precise than floating-point, leading to noticeable inaccuracies in the positioning of vertices and textures. This is why you might see textures “swimming” or vertices “wobbling” as the camera moves23. These inaccuracies were particularly evident in 3D models and environments, contributing to the distinctive look of PS1 games."

it's like if you try to have massive distances from origin, it starts to lose accuracy

[ornate hound]

https://www.youtube.com/watch?v=XWMPEE8O05c&t=82s

how can i recreate this ?

i don't understand what he does in the video

[proud hemlock]

Is there a way to create ambient occlusion using shadergraph?

[grizzled bolt]

No, shader graphs operate only per-vertex and per-fragment
AO by its nature requires calculations spanning more than one fragment/pixel

[grizzled bolt]

Post only to one relevant channel please

[ornate hound]

the video is about shaders

[kind juniper]

What about screen space ambient occlusion?

And screen space shader graphs in general.

[grizzled bolt]

I thought it had to rely on compute shaders to properly use the surrounding pixels for the distance and blur calculations, but I guess it could be done

Don't you need a whole new texture sample for every point of the depth and color textures which you'd read in a full screen shader

I can't confidently say

[kind juniper]

I think ddx/ddy is often used for that. It can give you data of the pixels in the same quad/fragment(or whatever it's called)

[grizzled bolt]

It is, but it's not nearly big enough for AO, or hardly any blur effect either

[kind juniper]

How about just sampling the positions and the normals of the surrounding pixels in a radius and if there are some that have normals different enough from current pixel, darken the color?

[grizzled bolt]

I guess my question is if sampling the screen texture multiple times for a blur or proximity effect in a fragment shader is more expensive / "worse" than reading it in a compute shader

[amber saffron]

It could work, but to have smooth and good looking AO, you need to sample a lot of pixels if you do it in a single pass.

[kind juniper]

Good question

[kind juniper]

I see.

[grizzled bolt]

Afaik texture sampling is the only way to do it in just one fragment, and it's preferable to have few texture samples in a shader

[amber saffron]

No, perf is about the same.
But SSAO relies on multiple downsample blur passes to work, then applying the effect is lighter by using those blurred buffers instead of doing a single pass blur with a lot of samples per pixel

[grizzled bolt]

I guess downsampling has to be done outside of the shader graph, but the rest of it could be done inside?

[low lichen]

If you mean still doing the blurring in shader on the downsampled screen texture, then that's still more expensive than doing the blurring outside and just doing a final sample in the shader.

[warm moss]

Who would even know enough to implement AO in shader graph but not be able to write some simple HLSL instead 🤔

[low lichen]

We're assuming they meant screen-space real-time AO. AO can also refer to AO textures, or some special cases where you can simplify it, like many voxel/Minecraft-esque renderers do.

but they probably meant screen-space ao

[grizzled bolt]

Probably ^^
I think all render pipelines have SSAO to begin with though

And if you were making one from scratch, compute shaders are probably the better tool for it

[amber saffron]

It doesn't necessarly have to be compute. Downsamling and AO can be done with regular pixel shader passes.

[tacit parcel]

SSAO is basically bloom, but instead of filtering bright pixels, we filter for creases (by comparing neighboring normals)

[strange basalt]

well they are the same as in they both are post processors in screen space

but finding all the bright pixels, bluring them then adding it on top does not really feel the same as using normals to find creases

[shell wadi]

Nothing wrong with good ol' fashioned vert/frag AO. Keijiro has an old (but still great) SSAO effect on his github

But doing it in shadergraph you'll run into the problem that shadergraph breaks (at least in HDRP) when you try to throw any mildly complex fullscreen post-processing at it

[deft gorge]

Does anyone have any idea why the HD Scene color is incorrect?

Applied to a plane there in front of a character to show it off

Am I missing something?

[grizzled bolt]

It looks like this is a Lit shader and scene color is being applied to base color, so your material will be affected by scene lighting

[deft gorge]

Yeah I need that for a frosted glass effect

[kind juniper]

What if you uncheck exposure?

[deft gorge]

On an unlit shader it's also slightly off

[deft gorge]

It ignores scene exposure haha

[kind juniper]

And unlit

[deft gorge]

What do you mean?

[kind juniper]

If you use an unlit shader with exposure unchecked

[kind juniper]

I see.

[deft gorge]

Yeah the exposure is needed

And I need the lit shader too because it needs to be affected by scene light

That is all expected

[shell wadi]

HD Scene Colour = pre-PostFX

So it will not have tonemapping etc on it which is worth considering

[deft gorge]

That must be it, it does indeed look like the tonemapping is off axis

Any way around that, seems like an odd limitation?

[shell wadi]

It's really not a buffer that's meant for 1:1 replacing the scene colour, it's more of a "useful for stuff that can look a tiny bit different" buffer used for stuff like bloom, some blurred refractions, etc.

Also it's broken half the time (in FS PostProcessing Shader Graph it doesn't work) but that's a separate story

[deft gorge]

In another project I got it close by color correcting in the shader, but that does not seem like the most optimal way to me

Do you have a suggestion for a better way to do a glassy refraction?

[shell wadi]

Only other thing to mention is try ensure your shader graph is using the correct precision - that might also be affecting it (if it's trying to render a fixed precision when the rest of the pipeline/scene colour then should still be float prec.)

[deft gorge]

Yeah the precision is set correct

Very annoying 😦

I guess I'll manually try to correct it a little again

It's not the tonemapping though, this is with tonemapping off

It's like HD Scene color uses it's own limited tonemapping

Whatever

Close enough for anyone who cares

[deft gorge]

coming from this*

[fair jackal]

So I have what are essentially 2D sprites that I'm implementing in 3D. For simplicity's sake, I want to have my normal maps for my objects represent their 2D perspective(as though I'm applying a normal map to a 2D sprite or a flat quad). However, my objects are going to be applied to a 3d mesh(like a cube), and thus my normal map won't work out of the box. How would I go about achieving this with a custom shader graph? I currently have an unlit shader, and I have a custom function for recreating the lighting, as I essentially want to do all of my object lighting with normals instead

basically I am hand-making all of my normal maps and I don't want to try to take into consideration how the normals would change on a per-face basis when I'm pixeling them

[amber saffron]

Set the shadergraph to use Object Space normals 🙂

[amber saffron]

You'll also want to export your texture as default without sRGB .
And sample them as "default" textures.

[fair jackal]

ah that seems a lot closer to what I was going for. Thanks! Though I am aiming for an unlit shader as I want to disable default object shading completely(as it conflicts with the shading that the normal map is doing) so that option is (obviously) gone. I have a custom lighting node function but I can't seem to wrap my head around the transformations. My results are close but don't quite look right?

[crisp python]

Hello everyone, i have been looking far and wide for this typ of shader in the glasses. Would anyof you know by any chance? When looking head on its okay but on the side its almost an RGB distortion?

[amber saffron]

1. According to the red-ish tint of the texture, You are still importing it as a normal map, but you should import it as a default 2D texture, with SRGB disabled.
2. Between the sample and normalize nodes, put a "remap" node, that remaps from the [0;1] to [-1;1] range. Since RGB colors can't be negative, but normals XYZ can, that's how it is unpacked.

[amber saffron]

Looks like a screen space refraction shader, Doing 3 refractions taps for each RGB component, with slighly different refraction indices.

[crisp python]

Thank you, now that I have the name I'll look for it, and if not I could try make it

[hearty obsidian]

Sorry I somehow missed that. I also ended up with what I believe is the same solution as you. I derived a class from the image component, overrid OnPopulateMesh and baked in linear uvs in uv1.

[strange basalt]

Yeah sounds similar, my is applied as a effect though so still using the original image

It surprising it's not like this by default

Things look so much better when you can 9 slice for nice crisp edges than add gradients or pattern

[hearty obsidian]

Agreed. I'm also not sure why this isn't default. The overhead of adding linear uvs to uv1 within the image class would be so little.

If you get a sec, would you mind sharing so I can compare the pros and cons?
Here's mine
https://pastebin.com/rdCfaNPb
It's not great cause it needs to repopulate the UIVertices but at least it retains the optimized mesh

[strange basalt]

Yeah give me a bit, will have to boot up my work machine

[hearty obsidian]

Thank you, no rush I come by every once in a while, I'll read you then if now isn't a good time

[strange basalt]

https://pastebin.com/C3pLgLhr

more or less exact same approach as yours, just doing it via BaseMeshEffect instead of deriving image

[hearty obsidian]

I like yours much better. The extra image component is a bit of an annoyance. Would you mind if I used your code?

[strange basalt]

go for it

also added the uv channel selection in case it ends up on a component that already does stuff with uvs

[hearty obsidian]

Not sure if you care, and it's not really a big deal, but GetUIVertexStream gives you the unoptimized mesh, so if you did this in your code

print(helper.currentVertCount);
helper.GetUIVertexStream(verts);
print(verts.Count);

You'd see a discrepancy in vertices count. You end up with the unoptimized verts. The solution I found was to manually populate them 1 by 1 to retain the optimized mesh.

[hearty obsidian]

Thank you very much

[strange basalt]

good to know, never noticed that but would explain a few weird edge cases

though not sure it matters since i manually populate when doing the actual modifications, the first loop is just getting my min and max so i can decide where the 0 to 1 space is

[hearty obsidian]

Hah, you are absolutely correct. I'll see myself out now

Wait no

You still populate in function of helper's length, and helper's length is modified when you fetch the vertex stream

[strange basalt]

but yeah good luck, it helped alot for getting the look i want, and yeah i will make the looking for the width and height consistient with the rest

[hearty obsidian]

Wait nvm I'm dumb

[strange basalt]

really will not say its optimized, i more or less just remebered this BaseMeshEffect class from before TMP exisited since its how you added shadows back then and just went for it with the api i saw without much actual reasearch

[hearty obsidian]

It's more than fine for my needs. I just needed glossy light shines on 9 sliced buttons and the likes. It's super handy to have. Again, thank you very much

[strange basalt]

yeah i use it for the same purpose, some shaders using it directly to just more or less add the V coord back on top of things to add a gradient

then a few shaders use it to map a actaul image based on the 0 to 1

like a circular shaped highlight on UI headers etc, or a really subtle pattern

just find it odd unity does not do this by default, i have worked on a inhouse engine before and i more or less added this by defualt when i added 9 slice support

[hearty obsidian]

absolutely, there is a clear need to span effects linearly across images in general. I agree

[broken vigil]

I need help with a Text Mesh Pro shader I'm working on. I've gotten pixel-perfect outlines working on Text Mesh Pro text, but the problem is the mesh itself that it's being drawn onto. Because the mesh is so tightly fitted to the letters, the outline does not draw correctly.

Extra Padding helps, but it isn't enough padding, and looking through the code of the TMP object, I can't find a way to make Extra Padding produce even more padding than normal.

Does anyone here know a way to increase the actual mesh size of TMP text, beyond what it normally uses?

[fluid peak]

not sure how useful it is to anyone else but I'm proud of making this for myself. to display a value in shader graph as an actual number. I wanted to interrogate what actual values were being passed back and forth, and this lets me look at what the actual numbers are. I'm sure there are fancier versions of this around but this one is mine...
https://drive.google.com/file/d/1ZnUGoF5LyvaKH7kSlOp_fKvruV-P1Y4d/view?usp=drive_link
https://drive.google.com/file/d/15fAuA0kQzslvKBMR-7QlaWeNgs1bIgYZ/view?usp=drive_link
https://drive.google.com/file/d/1L0mBmo8SUm_NCeext-ggQBFAvvtY77SK/view?usp=drive_link

[reef grove]

thats actually quite neat

for some reason the bounding box? of a light is affecting the color of my material

void finalColor (Input IN, SurfaceOutput o, inout fixed4 color)
{ 
    color = _DarkColor;
}

im using a final color thing, which has worked fine except for this issue. I can show the whole file if needed

but in theory a final color should be... the final color

Shader "Custom/Cell" {
    Properties {
        _MainTex ("Base Color", 2D) = "white" {}
        _BaseColorMultiplier ("Base Color Multiplier",Float) = 1
        _Normal ("Normal", 2D) = "white" {}
        _Dither ("Dither", 2D) = "white" {}
        _LightColor ("Light Color", Color) = (1,1,1)
        _DarkColor ("Dark Color", Color) = (0,0,0)
        _Bias ("Bias", Float) = 0.5
        _LightThreshold ("Light Threshold",Float) = 0.75
        _DarkThreshold ("Dark Threshold",Float) = 0.25
        _DistanceFog ("Distance Fog",Float) = 0.1
    }
    SubShader {
    Tags { "RenderType" = "Opaque" }
    CGPROGRAM
      #pragma surface surf SimpleLambert fullforwardshadows finalcolor:finalColor

      half4 LightingSimpleLambert (SurfaceOutput s, half3 lightDir, half atten) {
          half NdotL = dot (s.Normal, lightDir);
          half4 c;
          c.rgb = s.Albedo * _LightColor0.rgb * (NdotL * atten);
          c.a = s.Alpha;
          return c;
      }
    sampler2D _MainTex,_Normal,_Dither;
    float4 _LightColor,_DarkColor,_Dither_TexelSize;
    float _Bias,_LightThreshold,_DarkThreshold,_BaseColorMultiplier,_DistanceFog;

      struct Input {
          float2 uv_MainTex;
          float4 screenPos;
      };
      fixed4 _ColorTint;

    void finalColor (Input IN, SurfaceOutput o, inout fixed4 color)
      {

        
        color = _DarkColor;
      }
    
    
    void surf (Input IN, inout SurfaceOutput o) {
        o.Albedo = 1-((1-tex2D(_MainTex,IN.uv_MainTex))*_BaseColorMultiplier);
        o.Normal = UnpackNormal (tex2D (_Normal, IN.uv_MainTex) );
        
    }
    ENDCG
    }
    Fallback "Diffuse"
}

heres the whole shader

I changed the finalColor to be simpler but it did the same thing before (thats why it seems weird)

[reef grove]

this is odd

wait

I think I understand

its executing the final color for each light

so when the lights overlap

ohhhhh

thats weird

heres the thing

when I put nothing in the final color function it has no artifact

if I do color = float4(color.r,color.g,color.b,1) no artifact

but the moment I do color = color + 0.1 theres an artifact

but why would it make the color lighter if im just setting the color

color = float4(0,0.5,0.5,1) has artifacts

both forward and deferred have the artifact

im stumped

the issue is I can't find ANY documentation about the finalcolor thing

I can't even find where I originally learned it

https://docs.unity3d.com/Manual/SL-SurfaceShaderExamples.html

theres something here

its seems like its something to do with objects overlapping and the finalcolor being like

applying to previously rendered objects

but it happens with one object as well

something to do with lights and shadows

so it seems to be rendering twice but I still dont see how that would change things

if the final color is being set to a constant value

it draws for each light

which makes sense I guess

but what doesn't make sense is why the finalcolor is being... overridden? or something?