thick glen
Thank you all for your earlier help. Is there a way to read the component values of an object after a hit is detected with that object via the collider?
So something like
if(hit.collider.CompareTag("ObjectTag"))
{
//code that returns Component Value;
//code that returns variable value from script on the object
}
grand jolt
Hi guys, is there any way to add an IF check in here via nodes that basically wont feed in the emission map if it's null?
meager pelican
@grand jolt I'm not sure I tracked the whole conversation, so grain of salt.
But normally, you'd do
Now, #2 could be multiplying by your emission map's intensity value. So if you have black in the map, that's a 0, so you'd be adding 0. But if it's a 1, then you'd get full emission color x 1, and if .5, 1/2 emission intensity.
So finalcolor = saturate(baseColor + emissionColor * emissionIntensity);
grand jolt
you mean multiply it all after adding in the graph?
@meager pelican
i'm very new to this shader graph stuff, sorry lol
and yeah emission maps are black / white
have to also admit the emission is definately duller than expected, had to up the intensity by 4
meager pelican
No multiply has precedence over addition.
float4 emissionColor = [whatever you do] * emissionMapIntensityValue.
return saturate(baseColor + emissionColor);```
Assuming you're not doing HDR. If you are, lose the saturate operation.So in SG, make boxes that do that. :p
grand jolt
yeah i'm using HDR, so I have to find a node that will combine the RGBA of _MainTex and the intensity value of the _Emission?
If so, any idea how to get the intensity of _Emission seeing as it's an RGBA texture input?
@meager pelican
theres the entire map btw, in case i did something stupid somewhere
meager pelican
You said the emission maps are black/white. Right?
grand jolt
yeah
meager pelican
So just pick off the red value or something. It will range from 0 to 1, right?
grand jolt
ohh I see, theres only 2 values gotcha, so I can use pretty much any channel
meager pelican
And you have a color variable for emission color.
So if your RGB emission color is (.34, .23, .86) and you multiply by your black emission map color (what you called null in your question above) you'd be adding 0 to your other main color. So no change.
But if you have a 1 in your map, you're adding (.34, .23, .86) to the color.
Then there's alpha.....you'll have to deal with that separately, that's almost always a pain.
grand jolt
ah.....these sprites do have some alpha around them
meager pelican
Just use the alpha from the main texture, or alpha clip with a threshold.
grand jolt
this all started with a damn Brackeys tutorial, I just wanted emission maps on 2D sprites XD
oh ok, already done
ok, so:
RGBA of _MainTex
multiply with
R of _Emission
? assuming i'm not still being a total noob at understanding this
meager pelican
I love Brackeys, watch him every week just for fun, even if I know the topic.
I wish he'd learn more shaders. He's doing more SG though, so that's cool.
RGBA of main text.
Add (emission color * R of _EmissionIntensityMap)
grand jolt
the issue I have with brackeys is a lot of their stuff goes out of date fast, considering his following surely he could hire someone to go through and edit old videos with corrective subtitles or something
ok, trying now
@meager pelican - crosses fingers -
Look ok?
meager pelican
You're still going to have to do your vertex color thing too. That's all your "main color"
You did. 🙂
Did it work? Don't forget to save changes.
grand jolt
testing now, wanted to make sure it looked ok, before Unity fries my hard drive somehow 😛
it......works 
........according to what i'm seeing...
Does this solve the issue of _Emission being null? you mentioned it before but not sure if we addressed it with this
spiral agate
col += tex2D(Tex, UV - float2(d,0));
col += tex2D(Tex, UV + float2(d,0));
col += tex2D(Tex, UV - float2(0,d));
col += tex2D(Tex, UV + float2(0,d));
col += tex2D(Tex, UV - float2(d,d));
col += tex2D(Tex, UV + float2(d,d));
col += tex2D(Tex, UV - float2(d,-d));
col += tex2D(Tex, UV + float2(d,-d));
how do i put these into for loops?
first replacing 0->0, 1->d;
second replacing 0->(-d),1->(+d)
how would that look like in code
is radix4 possible in hlsl?
meager pelican
Black isn't null, it's zero. 😉
grand jolt
well yeah, sorry, forgive my noobishness 😛
spiral agate
Does this solve the issue of _Emission being null?
@grand jolt just clamp / saturate your output
meager pelican
He's using HDR
grand jolt
it works it's fine, HDR
spiral agate
alright then, max(youremission,0), at least fixes negatives
spiral agate
not sure, less code is nice to handle is all
meager pelican
Yeah, it might be. But really, a shader compiler is likely to unroll the loop if it can and it's reasonable size. Unless you tell it not to. They'd rather chew up code space than branch. IIRC.
spiral agate
alright, i'll keep it as is then
grand jolt
one last thing actually, to avoid coding blunders, some of my code will likely try to modify the "_Color" variable, would I just make a new color property as _Color, then feed that into where the Vertex Color is going in?
then i'll set it to white default
meager pelican
I thought you were using vertex color to maintain compatibility with the sprite rendering.
But you can do it with your own value if you want. _Color or _Foo or whatever.
grand jolt
possibly, and it all works fine, i'm currently using:
mySpriteRenderer.color = Color.Lerp(mySpriteRenderer.color, discoColors[discoColorsIndex], t);
But i just wanted to see if I can make sure if I ever forget that and try using the:
mySpriteRenderer.material.color = Color.Lerp("_Color", mySpriteRenderer.color, discoColors[discoColorsIndex], t);
way it wont freak out in the console
meager pelican
Do you have a lot of these that need custom [Disco!] colors?
grand jolt
yeah virtually everything, the script is per-object at the moment
the arrays are set in start
meager pelican
OK, here's the deal. It's a unity thing with materials.
If you set material.whatever it CLONES THE MATERIAL, makes an internal copy, and assigns i to renderer.material. So if you have 300 objects, you'll end up with 300 damn copies of the material, and that chews up memory, and screws up GPU instancing in the standard pipeline.
In the new pipelines, it's different, sorts by shader as I understand it, so it might work out, but normally you don't want 300 stupid copies of the same thing save one value.
So what you do is use Material Property Blocks. And you set the color on that.
Normally.
For frigg'n sprites, IDK for sure though.
But whenever I see .materail, I freek out.
Mostly you want to use MPB's and renderer.sharedMaterial. But in the new SRP's that can break one of the batching routines for the experimental renderer.
grand jolt
well i've got a bunch of test objects in the scene and it doesn't seem to freak out the batching at all
it all instances fine
also, I have a BUNCH of performance coding in it
meager pelican
Yeah, it might not, but it WILL chew up memory.
grand jolt
basically:
If you cant see it
If the component doesnt exist
If it's a certain distance away from the player
If the relevant bools arent selected on the script
It wont even act 😛
meager pelican
Like I say, IDK about sprites in SPR yet. It all depends on how many is many.
grand jolt
aye, even I dont really know yet, but i'm also planning to multithread and entity system eventually so hopefully that will mitigate any problems
DOTS is a beast..
also object pooling
this thing is going to be performance-optimized, no friggin way im ending up on someones dirty devs list
meager pelican
Indeed. I'm exuding optimism.
Yeah, pooling virtually always good.
Goodnight all. 🙂
grand jolt
thanks for all the help!
wary horizon
I have this shader effect, Would it cause alot of performance issues with it growing it size like that?
exotic estuary
I have a property exposed in my scene, and I can change the value in the shadergraph, but when I try to do it in the material inspector nothing changes. Anyone know where I might be going wrong?
(The property is an opacity variable and I'm pretty new to this fyi)
slow bear
What's cheaper for noise: a texture lookup or a function?
(in a situation where the shape of the noise is not important)
meager pelican
@slow bear Usually math is cheaper, but with low end mobile and if you're processing something pretty logically texture reads might be faster in some special instances, as low end mobile math can suck too, particularly when throttled due to things like heat or batter. IIRC. In general, favor math over memory reads. This is off the top of my head, and I'm not really up to speed so YMMV.
exotic estuary
Ok new question cause I sorted the old thing. How should I go about changing shader properties in runtime, and can I change them individually for each object?
I want objects to get transparent if they're close to the camera
Seems like quite a pain
meager pelican
You can have individual instance properties with Material Property Blocks, or by duplicating the material.
But for a nearness transparency, if all the objects share a material, then all you need to do is compute the distance to the camera position. The "shader" knows the camera pos, so you just use the worldPos of the frag and compute the distance. If it's < [some value] then go transparent.
exotic estuary
Right that would be ok except I want to use dither transparency
So theres no sorting hassle
So I'm trying to have one value for one object
By duplicating do you mean one material for every single object?
civic pike
hi there dear devs! has anybody an idea how I could create a mask for the scene fog in URP? I mean I would like to have a shader which is transparent and removes the fog on whatever will be rendered behind a mask f.e. thanks for any hints & tricks!
low lichen
@civic pike Can you explain what this is for? I'm having a hard time understanding exactly how this mask would work and what it would be useful for.
civic pike
imagine you are in a submarine for example. and you see the interior from outside. the interior should not be affected by the heavy fog which is surrounded outside the submarine
by seeing the interior from outside I mean that there is kind of a puppet house cut of the submarine, we are not looking through a window into it
low lichen
I see. Are you using the fog setting in the Lighting Settings window or a post processing fog?
civic pike
from the Lighting settings window
low lichen
Then the way the fog works is each object's shader is responsible for drawing fog in it. So if you make a custom shader that just doesn't do that, objects using that shader won't have fog on them
civic pike
I see, seems to be quite a task. Maybe I'm better off with a post processing fog.
What are the big disadvantages of a post processed fog, besides the performance costs?
low lichen
Well, post processing fog is applied globally and unless you customize the shader effect, you can't have some objects without fog
Unless you draw the outside environment first, then do the fog post processing on it, then draw the submarine interior on top of that.
civic pike
Yap, I think I get it 🙂 Thanks a lot for the quick response! ❤️
low lichen
I don't think it's unreasonable to have a custom shader for all the underwater stuff. Then you can also do things like desaturate colors/color overlay, which happens the deeper you go underwater.
https://i.insider.com/5dd69e19fd9db23c606b0e92?width=1100&format=jpeg&auto=webp
That shader would also add a custom fog effect, not reliant on the fog setting in the Lighting Settings window
civic pike
that's true, the point is that objects can be out or inside the submarine plus is the inner mesh the same mesh as the outer cover at the moment.
meager pelican
And if you're doing custom water and custom fog, you could use a stencil mask.
hollow grove
making custom volume from shader?
copper comet
I don't know if it's more of a shader or rendering question but I'm trying to use drawproceduralindirect, and it seems that the "start instance location" part of the bufferwithargs is not passed correctly to the shader ?
meager pelican
Huh. Standard pipeline?
What's happening?
Which function are you using (There's two signatures).
copper comet
yeah the standard pipeline
im using the first one (without the index buffer). I use the instanceid in the shader to get some drawing infos (textures, color etc) and everything is drawn using the infos at offset 0
maybe Im not understanding it corretly and start vertex position doesnt mean offset ?
(although it seems to work with the start vertex position but not the start instance position now that im doing some more troubleshooting)
meager pelican
God, it's been a while since I messed with this. They use the word "location"...but offset or index? 😉
copper comet
i would've hoped that the SV_VertexID and SV_InstanceID were modified accordingly 😛
meager pelican
I know I've had it working. Did you fix it?
copper comet
no, not the instanceid
im wondering if I need to use the unity macros
but its not really unity instance ?
meager pelican
There's macros to handle it all.
copper comet
no ?
meager pelican
Here's MS's HLSL explanation. And it's set by the API for you. But there's sometimes issues with how you declare it, so use the macros.
But yes ( 😉 ) it is the mesh instance.
And the vertexID will be the vert within the instance.
So you're going to have to do the math I guess. I'm trying to find you an example, but I'm juggling IRL stuff.
meager pelican
But the instanceID won't give you the whole offset that you need. I think you're going to have to multiply by # verts and the zero-based mesh index. Or whatever.
copper comet
yeah it must be something like that
meager pelican
So for a 3 vert mesh, you'll get 0, 1, 2 in the verts for instanceID 0, and 0, 1, 2 again for instanceID 1, etc. All zero based.
So your offset index into your buffer is instanceID * 3 + vertexID to get the vert data in your vert();
copper comet
but im using several call to drawprocedural indirection with different values in the argumentbuffer, using it like this (number of vertices,number of instance, vertices offset, instance offset)
instance offset doesn't seem to be taken into account ?
meager pelican
What is happening? How do you know "it's not taken into account"?
I mean, you may need some debugging here to see, like render doc to see what values are coming in, or even a gpu-specific debugger (AMD, NVIDIA gpu tools).
I don't know, I don't remember it not working last time I messed with this.
Location would be byte offset IIRC, as I'm sure you know. Per docs:
has to have four integer numbers at given argsOffset offset: vertex count per instance, instance count, start vertex location, and start instance location.
And that sounds like what you're doing!
Is vertex offset the "span" for each instance, or an offset within the instance? I don't remember.
copper comet
im just debugging as colors but yeah you're right its time to fire render doc
i thing the vertex offset is the same for all instances
think*
meager pelican
Yeah, but I'm assuming you have different meshes in one buffer, so you're making multiple calls, right? And location starts are call-specific.
But I just don't remember the diff...because I probably started ver0, instance 0 at 0 offset, so start vert location and start instance location was the same? My head hurts, and I have someone trying to talk to me IRL. lol
copper comet
ahah
yeah thats what i do.
hopefully things will get clearer with renderdoc ! thanks for the help
meager pelican
I didn't help much. Maybe when I get a chance I can find my old code where I messed with this and see what I did. But it was probably the same mesh all the way through. :/
Keep me posted if you figure it out, because now it's bugging me. 😉
copper comet
allright !
meager pelican
I guess the documentation could be a tad clearer on "start vertex location" and "start instance location" too. When would the instance 0 start at a different location than the vertex 0 of the first instance?
copper comet
hum apparently the two start location are not automatically set
slow bear
@meager pelican thanks for the reply; also, does favouring math over textures still holds true for Particle Systems?
copper comet
i dont know where i can access them
copper comet
yeah ! but the value i put in is not used anywhere
not matter what I set the result is the same
and renderdoc gives me the same values too et if i change the two offsets
copper comet
yeah
the 8 first int are 1920,1,0,0, 2880,1,1920,1
the code is a bit all over the place so there not a clear source i can show you
i draw two different mesh (they both have only one instance) one of 1920 vert at index 0, the other of 2880 vert at index 1920
meager pelican
You mean offset 1920?
That won't work unless you account for those 1st 8 ints. 😉
copper comet
and these are the calls
yeah offset 1920
the countbuffer is a different buffer than the buffer i read the vertex from
meager pelican
Oh, oops.
OK, if it's an offset, you have to multiply by sizeof item. You want a byte address.
Right?
meager pelican
So if your verts are float4's, that's 16 bytes each, yes? Because 32 bit float = 4 bytes each, x 4 floats.
Right.
copper comet
no i thinj the offset in the function is the offset in the count buffer
meager pelican
Dammit. I need to find some code.
Please stand by. One moment please.
copper comet
hum digging deeper i think either it's not at all how you should use that, or just it just doesnt work in this use case.
the actual d3d struct is this https://docs.microsoft.com/en-us/windows/win32/api/d3d11/ns-d3d11-d3d11_draw_instanced_indirect_args
meager pelican
Yeah, OK.
And Unity docs say this Buffer with arguments, bufferWithArgs, has to have four integer numbers at given argsOffset offset: vertex count per instance, instance count, start vertex location, and start instance location.
So "start vertex Location" and "start Instance location" would be 0 offset from the start of the item in your 1st case? Because you don't have any misc data at the start of the mesh.
And then you'd have for your 2nd one
A big offset for startInstanceLocation and 0 for start vertex location?
copper comet
i think id have start instancelocation at 1 and a big offset for start vertex no ?
meager pelican
Did you get the 1st instance to draw?
copper comet
tbh my brain is melting :p
yeah it draws
both call draw the same thing except the second one draw a big more vertex
meager pelican
The whole thing is float4's right? No extra stuff?
So maybe your 1920 or whatever.
copper comet
in countbuffer its int
meager pelican
I mean the vert buffer
copper comet
my "vertex buffer" its actually a structured buffer generated in a compute shader and there's a bit more stuff
meager pelican
Oy
copper comet
{
float4 pos : POSITION;
float3 normal : NORMAL;
float2 uv : TEXCOORD0;
float3 tangent : TANGENT;
};```OK OK OK 🙂
So the size of the item is 12 floats x 4 = 48 bytes (thinking out loud)
And you have 1920 in the first instance
Starting at pos 0 in the buffer
Then you have another >1920 in the next instance
copper comet
the next instance is actually a new call (just one instance per call just to test)
copper comet
im drawing the 1920 vert in one instance
im reading some data in my structured buffer, and some per instance data in an instance data Constant buffer
(with SV_instanceID)
meager pelican
Buffer with arguments, bufferWithArgs, has to have four integer numbers at given argsOffset offset: vertex count per instance, instance count, start vertex location, and start instance location.
You're setting up the arg buffer as
(1920, 1, 0, 0)
for the first call
copper comet
i would have hope the second call would have started the SV_vertexID at 1921 and the SV_instanceID at 1
yes
meager pelican
I would think so too
copper comet
right !!!
meager pelican
So for the 2nd call, what do you do that isn't working?
But since it's a 2nd call, it doesn't matter I suppose.
Come to think of it, I was expecting it to reset
copper comet
the second call draw the same vertices and per instance data than the first call
meager pelican
???
copper comet
even if the count buffer has the offset
meager pelican
how many verts in the 2nd call?
meager pelican
So you'd set argbuff with (2880, 1, ?, 0)
Wait!
Are you reusing the first set of verts, and just drawing MORE?
Or is it a completely different set?
copper comet
the arbuffer is like you said (with 1 in ?)
no
with 1920 sorry
im using the same structuredbuffer yeah
meager pelican
But not reusing the first set of verts. you have 1920 verts, and then 2880 verts. (your overlap comment confused me)
copper comet
yeah they are all in the same structuredbuffer
with the second mesh starting at index 1921
since i have an arbitrary number of meshes
meager pelican
So your second buff is (2880, 1, ?1920?, 0)? Maybe I have those last two backwards. sec
Because index 0 to index 1919 is your first set
1920 starts your 2nd set.
copper comet
yup
meager pelican
So your start Index Location (the last param) should be 1920?
If I read it properly (and IDK now)....
copper comet
i think its the one bfore last
meager pelican
So you'd set the buffer as
(2880, 1, 0, 1920)
Try that, but maybe I have it backwards. Spitballing.
The MS site says:
The index of the first vertex.
StartInstanceLocation
A value added to each index before reading per-instance data from a vertex buffer.```yeah i did that, to no avail 😦
meager pelican
And you tried the other way around. And that didn't work.
Hence your whole question here.
copper comet
yeah
im gonna try to see if i can ask someone at unity through my company contacts
this is mind boggling
meager pelican
I'm just extra stupid today, and I'm not trying it, I'm just trying to talk it through on a discord. ;)
I'll bet the forums would be more productive, if you post a question.
OH, maybe you have to add 32!
Because of your extra-structure data??????? And 0 offset for the vert, because it's the first structure member? Something like that
IDK Unity putting a pic in the docs, or more descriptive docs would help.
grand jolt
Hi Guys, got a problem with 2D sprite emission from yesterday, was hoping someone could help, The sprite is covered by emission is there if the sprite changes to a texture with no _Emission map
It would be great if I could add some sort of IF check in to make sure the _Emission is only passed if it exists....
(quite new, be gentle 😛 )
copper comet
@meager pelican yeah i agree. Il let you know when my contact answer
regal stag
@grand jolt If you change the _Emission map mode in the blackboard to "black" rather than white, it will be (0,0,0,0) when not set so won't show anything.
grand jolt
@regal stag Firm handshake to you sir, many thanks!
Would you know any tutorials on modifying this so it reacts to realtime light? as sprites are lit, would like to make them as dark as their environment
"I keep seeing "make it a diffuse" in places but not sure if thats relevant to me in the graph method..
regal stag
I think you are in HDRP, right? If so, I'm not sure. I only know how to light sprites using URP's 2D renderer & Sprite Lit master.
grand jolt
damn, curse HDRP and it's graphical quality 😛
think I found one, let me try modifying
fail...gah
grand jolt
Anyone know how I can get my 2D Sprite shader to react to realtime lighting?
I found a tutorial at https://www.febucci.com/2018/07/sprite-diffuse-shader/ but not sure which graph would be the equivalent on 3D HDRP..
knotty juniper
why not try it?
copper comet
@meager pelican It seems that the vertex offset only work if you provide an indexbuffer (and its an offset into this indexbuffer) I guess per msdn doc that the instance offset only work when you provide a vertexbuffer but unity doesn't allow that ? Ill wait for my contact info, but I think i'm just going to pass 2 int in a materialpropertyblock which are going to be my offsets, and it'll work fine.
fast rose
has anyone found a GOOD water that works in URP and cheap for mobile?
finite sequoia
Just a quicky, is there any way to pass a Texture2DArray into a shadergraph?
meager pelican
@copper comet indexBuffer......
The first version has no index buffer, the 2nd one does. They map to two different calls in DX11. One isn't indexed, the other is. You're using the first non-indexed one IIRC. But it still has the params in question.
It might just be easier for you to have two buffers, since you're doing two calls anyway, and then just pass zeroes for the last two params.
Your first mesh drew fine.
I assume the index buffer is vert index list for shared verts. So there's the verts list, and the index (unity calls it triangles) list.
But you're not using that version.
bitter forge
quick optimization question, i'm looking to add a bevel to my models, is there an optimization or performance concern if i do this with shaders? whats faster / more optimized?
meager pelican
@finite sequoia Did you figure it out?
Check versions on the site, but there's a sample Texture2DArray node, so there's a datatype for it. I think I looked at the 9.0 SG, but that's probably not production ready yet. So check versions.
@bitter forge I know how to do that in an SDF, but with a regular mesh, it's harder, as you have to add geometry to pull it off, since you're dealing with flat-triangles and interpolation. If you want to do "lighting tricks" that don't actually change the mesh, but just the coloring (so you'll still have sharp "points" on the mesh like box corners) that's a manipulation of the surface normal for lighting calcs. Still hard.
Maybe someone knows of a decent way to "add bevel", but wow.
The fastest way is probably to add the bevel to the geometry, or use all SDF's, but you have to know what you're getting into. If you want it adjustable, maybe you can reverse it...add bevel, but figure out how to REMOVE BEVEL in a shader. IDK.
Shaders deal with polygons.
finite sequoia
@finite sequoia Did you figure it out?
Check versions on the site, but there's a sample Texture2DArray node, so there's a datatype for it. I think I looked at the 9.0 SG, but that's probably not production ready yet. So check versions.
@meager pelican I'm using the TextureArrayNode and I have it as an accessible property, I just don't know how to pass data into it 😂Material.SetTexture2DArray()doesn't seem to be in the documentation anywhere, and nor does the lambda expression I had before to pass a Texture2D in viaMaterial.SetTexture(). I'm using the latest versions of Unity and Shadergraph
meager pelican
This exists:
https://docs.unity3d.com/Packages/com.unity.shadergraph@7.2/manual/Sample-Texture-2D-Array-Node.html
The input type is Texture2DArray.
According to this:
https://docs.unity3d.com/Packages/com.unity.shadergraph@7.2/manual/Property-Types.html
There's a property type for Texture2DArray. That's distinct from Texture2D. So use the Texture2DArray.
Not the Texture2D
On the C# side, you call
sharedMaterial.setTexture("_MainText", my2DATexture);
and see:
https://docs.unity3d.com/ScriptReference/Texture2DArray-ctor.html
So I guess the short answer is mat.SetTexture works for both. 🙂
@finite sequoia
finite sequoia
Thank you for all of this! I'll give it a tackle and let you know if I have any issues 😊😊
You're the best @meager pelican
bitter forge
@meager pelican Thanks for that. For what i read its a pain to do in a shader since the "when to calculate" needs to be stored somewhere so I'm doing it in blender using the bevel modifier in the meantime, this will add topology but thats fine. might get away with baking.
umbral cargo
can I write custom code to use as a node in ShaderGraph that performs custom vertex displacement?
I think I can and then pass the results into exposed properties like this on the master node, correct?
meager pelican
can I write custom code to use as a node in ShaderGraph that performs custom vertex displacement?
@umbral cargo Yes
I think I can and then pass the results into exposed properties like this on the master node, correct?
Yes
(You should just try it!) 😉
But the normal and tangents may need to be recalced if you're not extruding/reducing along the existing normal. If so, buckle up.
umbral cargo
ok, I think I am
just doing basic displacement map of the y-axis on a flat plane sort of thing
meager pelican
Well then no, you're not using the same normals/tangents. You need to calc new ones.
Because the normal of the plane isn't the same as the normal of a "hill".
The normal of the plane is Vector3.up basically. But that's not true on the side of a "hill". 😉
What you'll end up doing in calculating 3 points near your vertex and trying to decide on what the normal is. Which sucks when the vert is a peak, and it will be. You should do some digging/googling.
The normal calc, once you mange to figure out what 3 points to use, isn't too bad. It's trying to figure out what 3 points to use that sucks. Since your shader only knows its one point that it's working on. But there's derrivitives, and there's other ways I guess. Like sampling your height map and making some kind of decisions.
There's got to be tuts and videos somewhere.
Then once you have the normal, you'll be able to calc the tangent.
umbral cargo
oh I see
originally I had everything unlit, but I do want it to be lit now, so I will need to do that
sounds fun 😰
meager pelican
It's been done, there's info, and past posts in this discord.
But I was thinking that as long as you're thinking about custom nodes (you don't need one just to use a height map, you can do that in regular SG notes) you might want to just output all 3 things....vert, normal, tan.
umbral cargo
Interesting. Wasn't aware there would be regular nodes that would do that.
finite sequoia
@meager pelican Sorry to bother you again, I've read through the articles you've sent me and from what I can tell everything I have done is correct. I am trying to blend textures together based on their heights from a float[,] heightmap. I'm in the process of switching everything over to URP and shadergraphs. The Texture2DArray does seem to be passed over from the TextureData.cs script; I had a couple of Debug.Log() thrown in to verify. I will throw you screenshots of the shadergraph as well as the original HLSL and .cginc file that I used for the custom function
The original Terrain Shader
const static int maxLayerCount = 8;
const static float epsilon = 1E-4;
int layerCount;
float3 baseColours[maxLayerCount];
float baseStartHeights[maxLayerCount];
float baseBlends[maxLayerCount];
float baseColourStrength[maxLayerCount];
float baseTextureScales[maxLayerCount];
float invLerp(float a, float b, float t)
{
return a + t * (b - a);
}
float3 triplanar(float3 worldPos, float scale, float3 blendAxes, Texture2DArray textures, SamplerState ss, int textureIndex) {
float3 scaledWorldPos = worldPos / scale;
float3 xProjection = SAMPLE_TEXTURE2D_ARRAY(textures, ss, float2(scaledWorldPos.y, scaledWorldPos.z), textureIndex) * blendAxes.x;
float3 yProjection = SAMPLE_TEXTURE2D_ARRAY(textures, ss, float2(scaledWorldPos.x, scaledWorldPos.z), textureIndex) * blendAxes.y;
float3 zProjection = SAMPLE_TEXTURE2D_ARRAY(textures, ss, float2(scaledWorldPos.x, scaledWorldPos.y), textureIndex) * blendAxes.z;
return xProjection + yProjection + zProjection;
}
void layer_terrain_float(float3 worldPos, float heightPercent, float3 worldNormal, Texture2DArray textures, SamplerState ss, int layerCount, out float3 albedo) {
float3 blendAxes = abs(worldNormal);
blendAxes /= blendAxes.x + blendAxes.y + blendAxes.z;
albedo = 0;
for (int i = 0; i < layerCount; i ++) {
float drawStrength = invLerp(-baseBlends[i]/2 - epsilon, baseBlends[i]/2, heightPercent - baseStartHeights[i]);
float3 baseColour = baseColours[i] * baseColourStrength[i];
float3 textureColour = triplanar(worldPos, baseTextureScales[i], blendAxes, textures, ss, i) * (1-baseColourStrength[i]);
albedo = albedo * (1-drawStrength) + (baseColour + textureColour) * drawStrength;
}
}
^The .cginc function
The ShaderGraph
Sorry, I know this is a lot to ask, but I am super stuck
meager pelican
That's asking me to digest a lot of code before breakfast!
Let's try some debugging first.
Somehow you're calcing "height" to decide what index to use?
Or you have a mask?
finite sequoia
Yeah, the height value is calculated and is the minimum and maximum values generated by the perlin noise
meager pelican
And you're passing that into the triplaner thing as the last param. With a 1-val?????
I think you want that to end up being the index to read from in the array, right? Tex[0], tex[1], tex[2]...basically.
finite sequoia
Yeah
meager pelican
So you should be able to output some values for debugging
So take whatever that calc is, ignore the rest, and output that into the RGB color value. If you want it greyscale, output it into all 3 RGB channels. You may have to scale it so it ends up being 0-1, so divide it by the max. What do you get for the scene result?
The most common/easiest way to debug a shader is to dump a scaled value into the color.
After that it gets more complex....frame debugger, render doc, or dedicated shader debuggers.
finite sequoia
So take whatever that calc is, ignore the rest, and output that into the RGB color value. If you want it greyscale, output it into all 3 RGB channels. You may have to scale it so it ends up being 0-1, so divide it by the max. What do you get for the scene result?
@meager pelican So how do I go about that, sorry?
Setting the albedo straight to the textureColour value gets me my rock texture
meager pelican
I don't have my head around your code yet. Not awake!
But you end up with something that decides what to read. It looks like you're looping and blending though, and I'm not sure what all it's doing deciding on the blend.
My head hurts now. 😉
Somehow you want to look at individual variables, like ss, or or a scaled color or whateer you're trying to examine
meager pelican
You kind of are! Just output the numeric value in a color. and throw that into albedo.
So if you want to check what ss is doing, output ss at some point into the color. And output that from the node into albedo.
You're hacking your own code, and bypassing most of it (well, ignoring).
In a "real shader" you'd just "return ss.xxx/50." or something. For debugging.
finite sequoia
How do I grab the SS?
meager pelican
In the code. within the custom function. It outputs albedo as a vector 3.
So set Albedo = ss.xxx/50 or whatever the calc is that makes sense.
finite sequoia
Ah lol
SamplerState doesn't have a .xxx , and I can't output directly into the albedo as you can't cast a SamplerState to a float3
meager pelican
Oh, sorry, wrong one. Not ss.
Some other variable that you want to examine.
I was thinking "i" index.
Or something.
Yeah, not the sampler state, bad example :p
Try just figuring out the height first.
finite sequoia
This code worked using just 'real shaders' though. This is the code I'm trying to replicate
void surf (Input IN, inout SurfaceOutputStandard o) {
float heightPercent = inverseLerp(minHeight,maxHeight, IN.worldPos.y);
float3 blendAxes = abs(IN.worldNormal);
blendAxes /= blendAxes.x + blendAxes.y + blendAxes.z;
for (int i = 0; i < layerCount; i++) {
float drawStrength = inverseLerp(-baseBlends[i]/2 - epsilon, baseBlends[i]/2, heightPercent - baseStartHeights[i]);
float3 baseColour = baseColours[i] * baseColourStrength[i];
float3 textureColour = triplanar(IN.worldPos, baseTextureScales[i], blendAxes, i) * (1-baseColourStrength[i]);
o.Albedo = o.Albedo * (1-drawStrength) + (baseColour+textureColour) * drawStrength;
}
}
Alright! So in THAT code, heightPercent is the 0-1 value of the 0 to 50 (or whatev) for the y pos. That's good.
IDK what you're doing with:
float drawStrength = invLerp(-baseBlends[i]/2 - epsilon, baseBlends[i]/2, heightPercent - baseStartHeights[i]); yet. But maybe that works.
Let's find out.
output that into albedo, and return.
so Albedo = drawStrength; return;
finite sequoia
I get pure blackness
(also I have no idea why the water is white in game but previews fine in the scene view but that's a problem for another day lmao)
meager pelican
You can use unity's color picker and set some color somewhere (make a mat if you want) to the "pick" on any of those. You'll probably get 0's. Black. But you might get 0.00001 or something.
Just FYI.
OK, so your drawStrength calc seems to suck. At first glance. No offense.
finite sequoia
Hahahahah
meager pelican
So we have to figure that out.
finite sequoia
That's fine
Assertion failed on expression: '0 == m_CurrentBufferBindMask[kUnityPerDraw]'
UnityEngine.GUIUtility:ProcessEvent(Int32, IntPtr, Boolean&)
``` is a new errorI think you had some issues in your video too, and that could also be a problem. But we're debugging here, we expect problems.
finite sequoia
Haha
meager pelican
But IDK what that unity error is right now without googling for it.
Let's make sure everything is set in the 2dArray.
finite sequoia
It's a pretty common Unity bug apparently lmao, it just crops up from time to time
Let's make sure everything is set in the 2dArray.
@meager pelican Okey dokey. The array length gets set correctly, and there is data there. The files get accessed properly and the Tex2DArray has contents
Debug.Log(texturesArray.depth); gives the correct depth of 6
meager pelican
yeah, in c#. But there's a long journey from there to the shader settings and outputting results.
Just output
albedo = SAMPLE_TEXTURE2D_ARRAY(textures, ss, float2(scaledWorldPos.y, scaledWorldPos.z), 3); It won't look right but I just picked "3". Or some damn thing. This triplaner thing is throwing it all off a bit, but we have to deal with that.
BTW, that float2 UV setting looks "funny".
finite sequoia
By all means correct me lmao
meager pelican
THat's not what I mean.
I mean in what I typed. It needs to be scaled...the scaled worldpos.
So if it is, it's fine.
But include that calc first to scale it.
ANd then put a "return;" after it.
This takes a lot of time to describe in a discord! :p
finite sequoia
Including that line you sent applies the Grassy Rocks texture stretched along an axis
Changing the index changes the texture
meager pelican
OK, and that's the 3rd index (4th texture). So it's all in there.
Good. So we're sampling, and we're UVing something.
finite sequoia
Woooo 🍾🍾
We're living on a prayer
(because it seems like we're half way there.... It's okay, I'll walk myself out)
meager pelican
Yeah, I often crack myself up too. Not many others, but me, sure. 😉
OK, so let's check this out:
float3 baseColour = baseColours[i] * baseColourStrength[i];
float3 textureColour = triplanar(worldPos, baseTextureScales[i], blendAxes, textures, ss, i) * (1-baseColourStrength[i]);
albedo = albedo * (1-drawStrength) + (baseColour + textureColour) * drawStrength;```The problem is the darn thing is inside a loop.
What's your basecolorstrength[] ?
finite sequoia
It's actually not needed. Before I used textures, I used block colours to show the different heights. It's part of the Layer class and can be adjusted to tint the textures, but it's not essential
meager pelican
You're adding extra stuff and making my head hurt again! 😉
meager pelican
np
lol
OK, so drawStrength is still 0, but you invert that later in this loop. (1-drawStrength). Which makes no sense to me. But....maybe to you and I just don't understand what you're doing.
And I'm assuming you have baseColorStrength[] set to some non-zero values, right?
finite sequoia
the drawstrength is 0 if it is below the baseStartHeight for the index and 1 if it is above. This means that it is drawn at full opacity when it is between this layer and the next one and fades in from below
Yeah
Here is the data
finite sequoia
Hold on, it would probably help if I named everything the same lmao
meager pelican
lol
finite sequoia
Consistency 👉😎👉
meager pelican
P.S.
Normally you'd paste code with something like hastebin.
If it's more than a few lines
finite sequoia
It's all on github if that helps
unique cedar
Hello just wondering I use Unreal engine and Redshit and they are using the same metal/rough shading workflow
finite sequoia
unique cedar
Now for Unity wondering if it's the same workflow too? it been a very long time since I opened Unity
meager pelican
OK, got the github terrain shader open, reading.
finite sequoia
I created a new branch called Convert to Shadergraph that has the current stuff we are working on in
meager pelican
OK, found the cginc custom node for terrain. Not much to it, looks worse in discord. :p
finite sequoia
hahahah
meager pelican
Shouldn't you have some += inside your loop or something? Aren't you accumulating the color?
finite sequoia
Ummm probably?
meager pelican
You're multiplying in that last line. Maybe that's OK. Maybe not.
You're basically doing a *=
albedo = albedo * ____
finite sequoia
changing to += makes it white
finite sequoia
Looks like an iceberg
meager pelican
You can do a blend too. Blending it a bit more complex. It's adding, with the other value, but using alpha to weight it.
OK, better.
Let's blend.
Check this for fun. (sec).
https://docs.unity3d.com/Manual/SL-Blend.html
Now, that's NOT quite what we're doing, but it's a good read. What we're going to do is blend manually...not with the back buffer.
finite sequoia
Sounds interesting
meager pelican
so what you do is you blend in with alpha as an intensity (because that's what you're doing, an intensity). To blend, you can use several equations. But we'll try a few, and keep it simple. Additive blending maybe.
But weighted by strength.
finite sequoia
Okey dokey. What's the formula
meager pelican
Well, it looks like yours
albedo = albedo * (1-drawStrength) + textureColour * drawStrength;
if you imagine that the "draw strength" is the alpha.
I'm going to call drawStrength the apha.
so that's the very common alpha, one-minus-alpha blend. The more alpha you have, the less background you have. And vice versa.
finite sequoia
Yeah the drawStrenth is the alpha then
because it is how much opacity the texture has
meager pelican
Hmmmmmm
IDK why you got all white. Yet.
Can you maybe clone the material and make a test one, and then let's cut it down to two textures in the array (or 3 due to water).
finite sequoia
ofc
finite sequoia
Yup
meager pelican
And what do you get with the first 2 levels only?
finite sequoia
Thee changes have been committed to the branch
Black
Not solid black though, the normals are applied
meager pelican
You should really only need to change the mat.
finite sequoia
i know
meager pelican
OK
finite sequoia
But we can talk about my bad coding standards later
finite sequoia
Yup
finite sequoia
Sounds good to me
meager pelican
We're only looping through...how many times? 1 or 2 now? (There's the water level pink color thing going on)
meager pelican
OK, so 0 is water, and 1 is the mountains.
finite sequoia
0 up to 0.3 is the water texture and anything above is grass, yeah
finite sequoia
From the heightmap it's 1. There is then a multiplier that applies to the mesh generated
meager pelican
So scaled height
finite sequoia
yeah. Uniform scaling
meager pelican
Let's output the results on each variable
so inside this:
for (int i = 0; i < layerCount; i ++) {
float drawStrength = invLerp(-baseBlends[i]/2 - epsilon, baseBlends[i]/2, heightPercent - baseStartHeights[i]);
float3 baseColour = baseColours[i] * baseColourStrength[i];
float3 textureColour = triplanar(worldPos, baseTextureScales[i], blendAxes, textures, ss, i) * (1-baseColourStrength[i]);
albedo = albedo * (1-drawStrength) + textureColour * drawStrength;
}
Is what we're going to check out, when index == 1
meager pelican
So make a float3 debugResult outside the loop.
And assign it with
if (i == 1) debugResult = baseColour;
and then after the for loop, before the end of the function, just assign
albedo = debugResult;
finite sequoia
still black
meager pelican
Did you put the if after basecolor was calced (Sorry for asking that, no offense)
I assume you did.
So why is your base color black?
Is it supposed to be?
finite sequoia
It's meant to be sanddd
finite sequoia
the if statement is after the colour is assigned, eys
meager pelican
But actually, it should be the grass (1)
finite sequoia
And yeah I realised after I sent 😂😂😂😂
finite sequoia
It's black even if i == 0
meager pelican
So let's check each
finite sequoia
Okay
meager pelican
So set debugResult = to each part of the expression, and figure out what's zero and what isn't.
I'll BBIAB...
finite sequoia
bbiab?
meager pelican
be back in a bit
meager pelican
np. But everyone else is going to kill us for 2000 posts. 😉
meager pelican
IDK what they're doing in our discord anyway! 😉
finite sequoia
Hahahhaa
meager pelican
So did you figure out if the color or if the strength is 0?
finite sequoia
It's the strength. The textures are fine, and actually get blended, they're just on top of one another across the whole mesh
meager pelican
OK, well that's a start eh?
So maybe plug in a constant of .5 instead for strength, and use two textures.
And then we can look at the C# code that set the strength array later.
Do they blend?
finite sequoia
Back to black with 0.5
meager pelican
It's the strength.
@finite sequoia
Prove that then. Because it's not. ;)
Let's see, output the strength in debugResult.
finite sequoia
meager pelican
Well then why didn't .5 work?
Confused.
finite sequoia
I have no idea, that's why I'm here....
meager pelican
lol Fair enough.
Let's try .5 again.
So debugResult = baseColours[i] * .5 should work.
with the if still there of course
And the assignment of it to albedo at the end
finite sequoia
Now the mesh is the colour of the BaseColour of the Sand
finite sequoia
🎉🎉🍾
meager pelican
lol
finite sequoia
NOW we're living on a prayer
meager pelican
Yeah. And a wing.
So... It's that strength array not being set. And all colors are 0? Maybe?
finite sequoia
seems plausible
meager pelican
lemme go find your C#
finite sequoia
Carpe
You're gonna hate me
Idk if this is gonna fix it
but
tehere's a typo
Still broken
Phew
meager pelican
You gonna make me search all your code for "setfloat" or you gonna tell me where it is?
😉
finite sequoia
TextureData.cs line 31 was material.SetFloatArray("baseColourstrength") when it should have been material.SetFloatArray("baseColourstrengths")
meager pelican
Ah. That WILL do it. (cause it not to work)
finite sequoia
Still black
meager pelican
Debug log it and make sure you have the values there in the array.
Sec...
In your orginal shader, it was singular (baseColourStrength)
finite sequoia
Yeah
meager pelican
You see those 3 errors you're getting?
finite sequoia
they're not errors... They are what I am outputting with Debug.Log()
meager pelican
Uh, it's saying the sampler macro isn't declared. For some reason, although I guess it worked before.
finite sequoia
Ohhhh yeah that's a visudal studio thing I think
I can add the #define line if you think that will make a difference https://github.com/Unity-Technologies/Graphics/blob/master/com.unity.render-pipelines.core/ShaderLibrary/API/D3D11.hlsl
meager pelican
Up to you, if you think it's just VS and spurious.
But if not, you'll get black results if it nulls it out or something.
Or worse, nans (that's usually something else)
finite sequoia
No change, mesh is still black
meager pelican
OK, so if you output the results of the strengtharray, it was black. I mean REALLY black = 0?
OK, and no more errors and you know it all compiled.
finite sequoia
Outputting the strengtharray gives it the colour of the index
No more erros and everythign has compiled
meager pelican
It should be a grey value between 0 and 1, right?
finite sequoia
What should, sorry?
meager pelican
The values in the strength array.
So
if (i==1) debugResult = baseColourStrength[i];
Or plural version of that. But whatever.
And return debugResult in the albedo.
finite sequoia
We just get whatever colour is at the index
Ignore that
I'm an idiot
yes we get grey
meager pelican
OK, good. And if you're updating that per frame, or starting and playing again, it should change based on texture 1's slider value in the inspector.
So that's not it
Now anyway, maybe was before with the typo, I'm losing track.
finite sequoia
Haha sorry I know this is a lot
meager pelican
@finite sequoia Uhmmm....
Where do you use the value you calculate for baseColour?
finite sequoia
To add a tint; albedo *= (1-drawStrength) + (baseColour+textureColour) * drawStrength
meager pelican
OK, what I'm seeing on my (maybe out of date github) is
albedo = albedo * (1-drawStrength) + textureColour * drawStrength;
That makes sense
that you added it in
finite sequoia
The line above is as it is in the old shader o.Albedo = o.Albedo * (1-drawStrength) + (baseColour+textureColour) * drawStrength;
meager pelican
Yeah, that's how it should be I'd think. So whatever I'm looking at is a tad not-whats-going-on
fine
finite sequoia
I just pushed everythign I have atm
meager pelican
And you know you have the result commented out at the end, right?
meager pelican
It should be.
finite sequoia
Good then
meager pelican
Last I saw you had it at some value, like .3 or something, eh?
So that's the grey you're getting
So that's coming through.
finite sequoia
Yeha
meager pelican
Now just change your if to check the textureing results from the triplaner call.
finite sequoia
Wow I wish I could type
meager pelican
debugResult = textureColour;
finite sequoia
Oh I know, I was just talking about my typos
We have my grass texture mapped
So that works perfectly
meager pelican
And if you removed the debugging (comment out) it's black?
finite sequoia
Correct
meager pelican
All lines commented out, including the last one (Reeeally sorry for asking that, just can't see your screen)
finite sequoia
all the lines
finite sequoia
I added basecolour back in yeah
meager pelican
You had = not *=
you had albedo = albedo * (1- __) + (stuff) * ___
finite sequoia
Still black
finite sequoia
I'm as lost as you are by the soudns of it
meager pelican
You now owe me a beer, regardless of if I ever help figure it out.
meager pelican
Dang it! still black.
S'OK, slow day.
finite sequoia
I didn't think it would be THIS hard to transfer something I already had in HLSL to shadergraph
meager pelican
It shouldn't be, but maybe ...it could be me too. It's hard to pick up other people's code, I should fire up unity and load your github, but that's like cheating.
lol
finite sequoia
You haven't already?!?
meager pelican
Just on a wild hunch, change your metallic in SG to like .3 for fun.
No.
My goal was to help you do shader debugging, not for me to do it. Technically.
finite sequoia
Just to double check....
meager pelican
Still black when you run it?
finite sequoia
UnityEditor.ShaderGraph.Drawing.MaterialGraphEditWindow.IsDirty () (at Library/PackageCache/com.unity.shadergraph@8.0.1/Editor/Drawing/MaterialGraphEditWindow.cs:255)
UnityEditor.ShaderGraph.Drawing.MaterialGraphEditWindow.CheckForChanges () (at Library/PackageCache/com.unity.shadergraph@8.0.1/Editor/Drawing/MaterialGraphEditWindow.cs:265)
UnityEditor.ShaderGraph.ShaderGraphAssetPostProcessor.OnPostprocessAllAssets (System.String[] importedAssets, System.String[] deletedAssets, System.String[] movedAssets, System.String[] movedFromAssetPaths) (at Library/PackageCache/com.unity.shadergraph@8.0.1/Editor/Importers/ShaderGraphAssetPostProcessor.cs:66)
System.Reflection.MonoMethod.Invoke (System.Object obj, System.Reflection.BindingFlags invokeAttr, System.Reflection.Binder binder, System.Object[] parameters, System.Globalization.CultureInfo culture) (at <fb001e01371b4adca20013e0ac763896>:0)
Rethrow as TargetInvocationException: Exception has been thrown by the target of an invocation.
System.Reflection.MonoMethod.Invoke (System.Object obj, System.Reflection.BindingFlags invokeAttr, System.Reflection.Binder binder, System.Object[] parameters, System.Globalization.CultureInfo culture) (at <fb001e01371b4adca20013e0ac763896>:0)
System.Reflection.MethodBase.Invoke (System.Object obj, System.Object[] parameters) (at <fb001e01371b4adca20013e0ac763896>:0)
UnityEditor.AssetPostprocessingInternal.InvokeMethod (System.Reflection.MethodInfo method, System.Object[] args) (at <b85fd61302f04ebc817d5c985c202647>:0)
UnityEditor.AssetPostprocessingInternal.PostprocessAllAssets (System.String[] importedAssets, System.String[] addedAssets, System.String[] deletedAssets, System.String[] movedAssets, System.String[] movedFromPathAssets) (at <b85fd61302f04ebc817d5c985c202647>:0)
Yup and still black
meager pelican
OK wait a sec.
(searches dim recesses of memory)
you have a custom skybox, yes?
finite sequoia
Yeah I do
meager pelican
NO, that's not it.
finite sequoia
I can turn it off for now
meager pelican
I was thinking about something where skybox being messed up and not set in lighting tab turned everything black. but for you that's not it, because we can return a value in the shader, and it works.
What the hell is with all those errors?
finite sequoia
Meh, I'll use the default skybox just in case
Fuck knows
I've never seen them before
Shadergraph window stuff
meager pelican
If you keep getting errors save your work, including the SG, and exit and restart the damn thing.
meager pelican
That's the spirit! no half measures! 😉
meager pelican
lol
finite sequoia
Guess what? Still black
meager pelican
Errors?
finite sequoia
none
ancient kayak
Hello everyone! Do you know how i'd make a shader that takes the on screen position of an object and uses tht ro reference a texture? I want these asteroids to match the background
finite sequoia
@meager pelican dm?
meager pelican
Nah.
@ancient kayak what pipeline
And I don't understand your question, really. You can get screen position and map that to uv's to sample another texture.
ancient kayak
uh, universal i think lol
the question basically is - how do i make the asteroids take on the color of the background
ancient kayak
not quite either. essentially i want to recreate a foggy atmosphere, but using fog (to my knowledge) only makes everything one color
vs my backgrounds are gradients
meager pelican
That's different.
you want a texture-color based fog.
There's different ways to do fog.
You can apply it to each object...every object would have to be fog-aware. But you could always get that fog color from sampling a bg texture in screenspace. Then, as you set the color, you'd blend it with some fog color, maybe by distance.
If you do that, check unity's standard olde fog methodology.
But you can also do it as a post-process. That involves recoloring the screen, and applying distance based fog, possibly optionally maybe by checking the depth buffer so it's not at a far plane, or maybe by having a stencil mask for all objects drawn. or just all pixels.
That's done with a post processing pass and a full screen quad or triangle.
ancient kayak
😵
meager pelican
Start here
https://docs.unity3d.com/Packages/com.unity.shadergraph@6.9/manual/Fog-Node.html
and then figure out how to set the color from here:
https://docs.unity3d.com/Packages/com.unity.shadergraph@6.9/manual/Scene-Color-Node.html
ancient kayak
Give me a sec to process haha
meager pelican
But you'll have to do the asteroids in a later pass. Or maybe scrap that scene color node, and just use screenspace to sample your bg texture to get the color
ancient kayak
That's different.
you want a texture-color based fog.There's different ways to do fog.
You can apply it to each object...every object would have to be fog-aware. But you could always get that fog color from sampling a bg texture in screenspace. Then, as you set the color, you'd blend it with some fog color, maybe by distance.If you do that, check unity's standard olde fog methodology.
I think I've seen "by distance" before, but that makes things change color the further away they get from the camera, right? I need things to change color based on the xy.
meager pelican
The distance isn't about the color, it's about the intensity of the color (more/deeper as it is farther away, less up close)
ancient kayak
Hmm. I'm not sure I follow/if we're on the same page. If all the bg asteroids are the same distance, how would we make them different colors?
meager pelican
You said you want to tint them toward the bg color. So pixel by pixel, you'd sample the background color at that point in the screen, and use that color. But how intense the fog is ...that's up to you to decide how to do. Often done by distance, maybe with a scalar so you can set intensity.
ancient kayak
Ah I see
meager pelican
If they're all the same distance (that's doubtful for a 3d object) then maybe just scale it.
ancient kayak
I guess now I have to figure out how to sample a screenspace texture
In case it wasn't clear, I'm a total noob with shaders lol
meager pelican
There's a node where you can get screenspace location. ANd then you can sample your bg texture if you can convert it to UV's (0-1 values for x and y)
google around and play with it, now that you have the general concept in your head. You'll get it. 🙂
meager pelican
np!
finite sequoia
It's okay @ancient kayak I know your pain
meager pelican
@finite sequoia Tell me you were debugging and you fix it while we were talking. 😉
finite sequoia
Haha nope, I went and had some dinner lmao
meager pelican
They let you eat with code like that? 😄
OK, so I'm going nuts. Let's check the albedo outputs for index == 0
So in the if() check if i==0 and then save off albedo into debugResult, and output debugResult;
meager pelican
Could be, I think we got zeros. That's where we're going next
In that albedo calc.
finite sequoia
Hold on
This is what I had before...
float invLerp(a, b, t){
return a + t * (b - a);
}
And I have now googled what the formula actually is
float invLerp(float xx, float yy, float t)
{
return (t - a) / (b - a);
}
Changing the function gives a funky result...
meager pelican
I was going to bypass it next anyway
finite sequoia
meager pelican
Better I guess. But hang on
You know where you declare float3 albedo, and then set it to 0? Set it to .5
meager pelican
OK now before your albedo calc, add a debug line and set drawStrength = .5;
finite sequoia
Like this?
if (i == 1) {
drawStrength = 0.5;
}
albedo = albedo * (1 - drawStrength) + (baseColour + textureColour) * drawStrength;
Nah.
Just maybe right after the inverserLerp line. Say "drawStrength = .5;"
Bypass the whole inverselerp thing
finite sequoia
Looks a bit better
meager pelican
I can't tell...did you remove the debug lines except for the new one?
meager pelican
So it should now be blending each texture 50/50
finite sequoia
i think it is, yeah
meager pelican
What was happening before was that you started off with black, got 0 for your inverseLerp so you blended 0 with 0's original color
A lot
meager pelican
yep
So now you know A) how to debug this, even if the long way around, and also where to fix it.
Sec...I'm working on lunch for someone
finite sequoia
Okay!
meager pelican
OK, so way back you said this was the original source:
void surf (Input IN, inout SurfaceOutputStandard o) {
float heightPercent = inverseLerp(minHeight,maxHeight, IN.worldPos.y);
float3 blendAxes = abs(IN.worldNormal);
blendAxes /= blendAxes.x + blendAxes.y + blendAxes.z;
for (int i = 0; i < layerCount; i++) {
float drawStrength = inverseLerp(-baseBlends[i]/2 - epsilon, baseBlends[i]/2, heightPercent - baseStartHeights[i]);
float3 baseColour = baseColours[i] * baseColourStrength[i];
float3 textureColour = triplanar(IN.worldPos, baseTextureScales[i], blendAxes, i) * (1-baseColourStrength[i]);
o.Albedo = o.Albedo * (1-drawStrength) + (baseColour+textureColour) * drawStrength;
}
}```Yup it is
meager pelican
So what that's doing is giving you back "t" for whatever value you have in worldpos.y
finite sequoia
Soudns right
meager pelican
What you're doing is
float drawStrength = invLerp(-baseBlends[i]/2 - epsilon, baseBlends[i]/2, heightPercent - baseStartHeights[i]);
finite sequoia
yup
meager pelican
and I'm not wrapping my head around that.
finite sequoia
Carpe I fixed it
We needed to saturate the result of the inverseLerp
float invLerp(float xx, float yy, float value)
{
return saturate((value - xx) / (yy - xx));
}
Thank you for all your help! Now let me buy you a virtual beer
finite sequoia
It's not a recognised symbol
meager pelican
huh.
meager pelican
maybe I'm thinking smoothstep or C# or something.
But OK, if you can fix yours, you're gold.
meager pelican
Well, between the typo on the strengths, and also the invLerp thing, it was a combo problem, and hard to figure out.
Glad to help.
finite sequoia
Any ideas right off the bat as to why the water doesn't appear in Game view?
meager pelican
Is it SG?
finite sequoia
Yup
meager pelican
Seems like others have discussed such. I'm not sure the context. But off the top of my head, no. Maybe show the graph so people can comment.
finite sequoia
It's a big boi. I'm gonna do some googling first and see if I can fix it on my own
Just wondering if you had come across anythign similar in your time
meager pelican
Yeah, divide it up into parts. Use the color output trick to check values.
finite sequoia
:))))))0
meager pelican
It's probably got reflections, but I see you replaced the skybox, so that's not it. Sometimes you can get NaN's from invalid skybox or something.
I'd bypass any reflection calcs first, and just see if it's blue. Etc.
Also make sure you bake any lighting that you need to.
finite sequoia
@grand jolt personally, I'm not a fan of the blooming, I'd also give it a more shallow depth of field. Other than that this looks amazing! What's it for?
willow summit
anybody know how i tell a object not to cast shadows while having a costume pbr graph shader?
clever plaza
Hi guys, do you know why my wave shader is different on scene and game view?
I want to achieve the scene view look
grand jolt
@finite sequoia that is a picture from my survival game , my job in this is i make the graphics/map/main menu/character and controls
finite sequoia
Awesome! If you need anyone to test, hmu! @grand jolt
meager pelican
@clever plaza Possibly due to
clever plaza
@meager pelican
I see, so i should calculate the vertex in worldSpace?
Yes different camera pos / changing ortho size causing different result as well
meager pelican
I would think that you want the results to be world-scale specific, regardless of where the camera is viewing it from. Thus the scale of the sin wave should use world-units, and THEN go through the perspective/matrix changes. So in the distance, the waves would be very small on the screen but the same # of waves across the object. Same way the object scales.
The vert() shader needs to output clip-space, so that's good. You need to keep that much of it. But somehow, you need worldspace in there too. You'd probably have a separate worldspace calc variable. There's a unity_ObjectToWorld matrix you can use.
Or if you're not scaling it and don't need to transform the object accordingly, you can modify it directly in object space, changing it BEFORE the conversion to clip space.
clever plaza
Ahh okay,
Thank you for the detailed explanation.
After modifying the vertex wave (in object space) and then convert it to clip space, now the scene and game view has the same result.
Thanks again 😁
meager pelican
Just be careful how the scale and rotation of the transform works, or doesn't. 😉 Depending on what you want. 🙂
Cool that you got it working.
frozen mason
Hey guys. I need help. Im looking for a way to sample texture in order to change the vertices of my procedural cube grid so it would give me forward motion for specific cubes based on that texture. I want to build something like this. https://www.youtube.com/watch?v=6wWD3kxV1og&feature=emb_title
What I have is the procedural grid and some shader that can move the z axis of the vertices so it gives me that effect but cant understand how do I sample the texture so only cubes that match black or whit color would move instead of whole mesh z axis
Here is my shader what I need to adjust to sample the textire and move vertices based on it
meager pelican
@frozen mason
But if you're adjusting verts (or position) in the vertex shader you'll need to use tex2Dlod to sample in a vertex shader stage. Then you can adjust either the position, or extrude verts along their local object z axis probably BEFORE you transform it to clip space.
frozen mason
@meager pelican URP. So with the above shader I can extrude verticec on z axis but I tottaly dont know how to incorporate texture sampling into that. So black will be 0 extrude and white will be full extrude
Im total noob with shaders.
so forexample in this gif it would be the dark grey would not move where the white grey only move
meager pelican
We have a Cyan for that. Wonder where he is? 😉
regal stag
Vertex offsetting involves adding or subtracting a value from the existing position. If you were to multiply that offset with the result of a texture sample before the add/subtract, it would give the result you want.
Assuming the texture is black (0) for no offset, and white (1) for full offset
And as Carpe mentioned, since it's in the vertex stage you need to use the tex2Dlod / Sample Texture LOD node, as the normal texture sample can only be connected during the fragment stages.
frozen mason
Yeah I know about LOD node from multiple articles. What I dont understand is how to properly connect the nodes to get this
If you were to multiply that offset with the result of a texture sample before the add/subtract, it would give the result you want.
Am I at least close to what u wrote?
regal stag
Yeah, I'd probably do the multiply after the Branch so it affects both the Left and Right extrudes at the same time, assuming you want it to.
frozen mason
tried that. Don't seem to give me the effect. Still extrudes all of the vertices of front ond back face
regal stag
You could also swap the "Comparison -> Branch with 0 and 1" from the texture sample for a Step node. I think they would be identical but makes it a bit more readable imo.
If I'm not mistaken, both the grey colours in the texture are over 0.5
Think they are around 0.62 and 0.78 from using the eye-dropper tool in paint
If possible I'd just switch it out for a custom black + white texture.
frozen mason
You could also swap the "Comparison -> Branch with 0 and 1" from the texture sample for a Step node. I think they would be identical but makes it a bit more readable imo.
@regal stag hehe not feeling what you mean. The comparision and branch is clear for me at least for now so I know whats happening somewhat
Alright something is happeing but it only affects 2 edge cubes and in a weird way 😄
regal stag
Might be that the UVs are too close to the transition on the texture. Maybe messed up by the linear filtering. Don't know if using point mode would fix it or not though.
Would be a setting on the texture in the inspector, or you could connect a Sample State node to the Sample Texture to override it
I think I'd probably use the position of the center of the cube to sample the texture (using the Object node), that way all four vertices will always be the same
Also is each of these black/white blocks a separate cube?
frozen mason
Yeah might be the UV i had a problem with setting them on my procedural cube grid. I thoguh I did them properly as compared to the unity default cube and they looked
the same.
Yeah it is
I;ve added Sampler state and set it to point filter and wrap repat and increased comparision threshold for the texture and this is the effect Xd
I think I'd probably use the position of the center of the cube to sample the texture (using the Object node), that way all four vertices will always be the same
@regal stag whoa that's something new. Truly don't know what ure talking about. Im really soryy being so hard to work with. I have 0 experience with shaders 😦 Apologise
regal stag
If you are generating the cubes procedurally you could set the UVs for all points in each cube to a single point in the texture. That way it would always be inside the black/white parts, rather than on the edge. Unless you actually need these UVs for applying a texture to each face too - in which case you could use two separate UV channels to pass this information into the shader.
I think you'd basically want each cube's UVs to point to a different pixel on the texture.
frozen mason
so the UV mapping gave me a lot of headache. THis is what I do for the texture
private void SetUvs() {
float tilePercentX = 1 / texturePixelSize.x;
float tilePercentY = 1 / texturePixelSize.y;
float uMin = tilePercentX * tileX;
float uMax = tilePercentX * (tileX + 1);
float vMin = tilePercentY * tileY;
float vMax = tilePercentY * (tileY + 1);
_uvs.AddRange(new[] {
new Vector2(uMin, vMin),
new Vector2(uMax, vMin),
new Vector2(uMax, vMax),
new Vector2(uMin, vMax),
});
}
So giving it tileX = 0 and tileY = 0 it would give me the bottom left square at least that what I think it should 😄
regal stag
Right, so I think if you did
float v = tilePercentY * (tileY + 0.5f);```
for all four vertices / _uvs, it would be a point in the center of the pixel for each cube/tile.Oh I get where you going. Yeah, make sense
regal stag
That way it's unlikely to bleed into other pixels and cause issues, and I see no point of the min-max unless it's bigger than a pixel and is used to apply an actual texture/pattern.
frozen mason
Fik ME!! IT WORKED
OMG @regal stag you are the god!!! Holy I was fighting with this for 2,5 weeks. Im so happy YEAAAAA
Holy moly... Fianlly. Thank you so much man! I love you and your blog. It's very helpful!
regal stag
No problem, and thanks! 💙
humble mantle
How do you use shuriken particle system with hdrp? My particles dont fade nor do they show the colors, the particles on show the colors from the sprites (which are just white)
I cant find anything helpful online either, so kinda stuck here
regal stag
I believe shuriken particles use vertex colours
humble mantle
is that not compatible with HDRP?
regal stag
Should be, the shader just needs to take it into account. If you are using shadergraph there's a Vertex Color node.
humble mantle
I am just using the default hdrp lit shader with shuriken, I assumed it would work like all previous default shaders Unity offered
regal stag
I don't think the default lit shaders tend to use vertex colours. Usually you'd use a Particles shader. I assume HDRP has a particle material somewhere
If not, could probably create one via shadergraph instead.
humble mantle
I am an absolute rookie with shadergraph, is CG still a thing with HDRP?
regal stag
Uh not really. I'm more familiar with URP but while you can write written vert/frag shaders there's a bunch of differences. It's a lot easier to use shadergraph.
humble mantle
Alright, cheers for the info, I will be spending some time on tutorials then ❤️
pure flume
Hey, does anyone know a good starting point in shadergraph for isolating certain parts of the frame to be rendered? Say I have a sphere and want everything thats inside that sphere to have a changed color or be blurred.
I'm trying to figure something out with camera position or view direction and calculating from there but I can't even begin to get my head around this.
fervent tinsel
@pure flume urp or hdrp?
pure flume
@fervent tinsel hdrp
fervent tinsel
you can use HDRP's custom passes for something like that
this has some examples https://github.com/alelievr/HDRP-Custom-Passes/
you could also use stencil mask but you also need custom pass atm to set the stencil bits and it's 100% undocumented
@pure flume check especially the glitch effect from that
it's using shader graph along with custom pass config for the effect
(no coding required for that)
pure flume
I had seen bits of that custom pass stuff, but didn't understand it well enough to determine if it was what I needed
I suppose getting the effect I want from a normal shader applied to a single object doesn't really make sense
I will read through it again and try to apply it. Thanks 🙂
fervent tinsel
if you just want to change the color of single object, you could just do that on one custom shader
no need for custom passes and stuff
pure flume
well I'm not sure if I'm explaining it correctly
what I want is more complicated than that though
fervent tinsel
well, can you explain what you really are after?
pure flume
I basically want special effects (not entirely decided yet, possibly greyscale or blurring etc.) applied to a specific volume, more or less
if I applied the shader to a box lets say, I would expect everythign inside that box to be rendered in this special way any anything out of it normally
or vice versa
*and
thinking about it, it does sound like a fullscreen pass
with some depth buffer magic
rather than a standard surface shader or anything like that
knotty juniper
The volume does refer to if the camera is inside that volume it uses the effect
not if the object rendert is
pure flume
sorry, I don't quite understand. are you asking a question or telling me?
knotty juniper
Telling
if you are refering the the post effect volumes
pure flume
I'm not really referring to anything specific tbh
Since I'm not entirely sure what it is I need
current theory seems to be Custom Passes
not post processing effects exactly
knotty juniper
Custom Passes sounds good
pure flume
Well, I guess this is a little above my paygrade for now 😄
obtuse dirge
Getting Openssl error trying to upgrade to HDRP 7.4.1 from 3.1 , Unity 2019.3.13f1, but all the requirements went fine
ON top of that , after trying to send in crash , 10 min later or so, said unable to write ,,,?
warped field
is there any shaders for a sketchy cartoon look
?
not just a outlike but like a scribbly shader
hexed portal
for anyone doing procedural terrain and wanting to avoid this kind of thing
just found a blending algorithm that seems useful
they pretty much store both textures' "depth maps" as grayscale versions in their alpha channels
fervent tinsel
@hexed portal looks like pretty stock standard height blend
HDRP's terrain shader supports that out of the box
I dunno about URP as I don't use it
hexed portal
yeah, I don't really need the features from HDRP/URP at the moment but this specific functionality was nice to find
fervent tinsel
I bet many 3rd party terrain shaders implement that as well
it's pretty common technique
hexed portal
I didn't actually know how it worked until reading that tbh
didn't know it was that simple
ancient kayak
@meager pelican i'm baaaack. idk if you're online though lol
does anyone know how to make fog affect individual elements differently? vs just having one global fog?
soft harness
So i'm trying to have this heart beat thing, but it's not exactly working. Any way to have this be a more heart beat like pattern? float beat = lerp(v.r, 1, sin(_Time * 50));
royal night
Any idea why would be getting the error
maximum ps_5_0 sampler register index
``` when I'm trying to modify a third party shader that is in a my assets?
The original is imported/compiles fine, but when I make a copy and move it into a different folder, it starts throwing that error.
burnt wigeon
@soft harness check LeanTween or DotTween
soft harness
What?
royal night
nevermind, I think I see what's happening, the folder I copied it into had some overrides for UnityPBSLighting.cginc and etc, and I think one of the parent includes has more samplers defined.
burnt wigeon
@soft harness Or animation curves. Tweening engines let you the stuff you wanna do with one line
you could pass an animation curve to your lerp too and tune it the way you want
meager pelican
@ancient kayak Well, you didn't define "differently", but we talked about two ways to do fog...on the objects themselves as you draw them, or with a post-process of the screen.
If you want each object to have different fog based on not just the background color, but some attribute of the object, I suppose the "do fog as you color the object" method would be most attractive. Then you can do whatever you want as you assign the color.
But I'm guessing here, as I'd can't really tell what you're after.
The "do it as you go' method has the advantage of not needing a 2nd pass of the screen in post.
But it's limited in some ways too. Like harder to do volumetrics and shadows in the fog...because it only knows about itself since the scene isn't constructed yet, and shadow passes happen later.
I suppose you can have a hybrid approach where the object writes to some buffer (maybe stencil) and then the post-process knows about that, and can use the customized info to do whatever you want.
signal rapids
Hey! Sounds like a simple thing, but I can't figure out how to create a warp effect like the warp node in Substance Designer. I want to push/push areas of my texture with a normal map or a noise. Can someone point me in the right direction? I'm using Shader Graph in HDRP
grand jolt
is there a way to transfer vertex paint data to other LODs? i realize it wont be exact because different meshes, i just need it more or less similar
meager pelican
@signal rapids So your UV sampling would be modified by some warp-calc, like noise. Done all the time, usually just adding some offsets to it. Remember that UV's range from 0 to 1 for x and y.
@grand jolt it has to be built into the model's vertex colors and/or UV's. If you're using procedural geometry, you'd calc that anyway.
grand jolt
@meager pelican so theres like no way to make every single object unique without having unique prefabs? i just wanna mix it with shaders later, LODs are for extra performance. should i use one LOD then?
fervent tinsel
use decals / shader with randomization that doesn't rely on vertex color
or simply use mesh without lods
meager pelican
LOD's are kind of debatable, they're good for far-off objects, otherwise it's mostly about on-screen pixel count, not triangles per se. Popping is a problem close up. One of the experts here had a post/article about it...I think it was Jason Booth.
Anyway, so if you want LOD's, and you only need one or two max, you'd have to paint them too. That's if you want painting that way with a tool. The other way is to just "paint" the texture that you've uv unwrapped into.
grand jolt
Hello I use DrawMesHInstanced instancing to draw thousand of objects on screen which used to work fine on all platforms.
some months ago after a chrome update the instancing on chrome is meshed up while on firefox/windows build everything is ok
any ideas?
fervent tinsel
ah yes, you can use "splatmap"
I think tools like substance painter has some built-in support for like 4 layer materials
but chances are that you'll pay more from texture lookup overhead than what you'd pay for simply having one lod
your mesh doesn't exactly shout highpoly to me
like with all perf related things, don't assume, measure
make some prototypes with estimated polycounts and see if it matters at all
meager pelican
Yep, measure.
But there's some givens. Like those hi-poly mountains you're walking on up close...if you use them for distant backgrounds...well, you don't want one triangle per pixel or worse multi-per-pixel, since you overload the vertex shader stage for no reason. So a lower-poly version is great...until you travel there.
brisk bison
I downloaded a transparent terrain shader, but it's having strange bugs on my terrain
I got it from here
If I make a new terrain and apply a transparent texture, it seems to work fine
So, why would it only have a problem with my current terrain?
Interesting, turning off GPU instancing seemed to stop the bug
fervent tinsel
@brisk bison I'm curious, why would anyone need transparent terrain?
brisk bison
Mainly because I need to cut out part of the terrain, and I can't find any way to do that
brisk bison
I'm using Unity 2018.4.20 sadly
fervent tinsel
I actually totally understand that
2018.4 is still more stable than 2019.4 IMO
and if you use built-in renderer, there's not all that much new on 2019
(better terrain though, and newer physics engine)
brisk bison
The terrain tools has me drooling, it's sad I can't use it :/
fervent tinsel
yeah, it's a nice improvement to the previous setup that was unchanged for a decade
just for holes though, you might get better perf with opaque shader that does alpha-cut
don't ask me how to do it for this though 😄
brisk bison
I'm trying to basically cut a plane (water) through the base of the land. For my use case, I need it completely see through to the underside of the world (absolutely nothing, just the skybox)
fervent tinsel
for that you could also just have multiple terrains that align so that there's a hole in the middle
brisk bison
So I figured , if I can't cut the terrain, made I can use a shader to render it transparent
fervent tinsel
but even that would be easier to do with newer terrain where you can adjust terrain across the seams
brisk bison
Yeah, the 2018 tools are a little limited with that kind of flexibility
fervent tinsel
for the terrain shader and gpu instancing... did you just copy paste some shader from internet or did you modify a 2018.4 terrain shader?
fervent tinsel
there are probably some changes on the terrain shader which are required for the gpu instancing to work
basically you can download 2018.4 shaders from https://unity3d.com/get-unity/download/archive if you want to mod from the real deal (which is what I always do if I need to change stock shaders)
when you hit the Downloads (Win) etc dropdown there, there's an option for Built in shaders for the version you select
shaders that ship with the editor are compiled so you can't realistically edit them as is, you need the source code which is on that mentioned link
brisk bison
Ahh. I bet that if I could write shaders, I could make it render transparent below a certain height
fervent tinsel
sure
that's actually not that difficult
I don't deal much with built-in renderer anymore but when I do, I usually start with Amplify Shader Editor (you could also use free Shader Forge) to see what code it generates so I don't have search around for right shader functions and macros
simplest way to do this type of thing is to take the vertex position and check if it's below your float value you set in some property, if so, make it fully transparent
doing it in object space would make it work regardless of the terrain's own location
brisk bison
I have Amplify actually, never used it though
Looks like my best bet here it to attempt to write one. Gotta learn something sometime I guess :p
open leaf
Hi, Unity 2020 beta exposes Conservative Rasterization flag - how can I enable it when doing custom shader rendering using command buffers?
willow pike
Hey ya'll! I've got a simple fire shader I made for some torches in my game. It's working great, but I'm wondering the simplest way I can tweak the shader so there is some randomization on a per object basis. Say I have two torches side by side, I don't want their animations to be identical at any one frame.
meager pelican
@open leaf Did you try using the SV_InnerCoverage semantic?
burnt wigeon
@willow pike you can instanciate a copy of the material and tune it
meager pelican
IDK if it's active or not, just reading the docs
burnt wigeon
copies dont share the same parameters
meager pelican
@willow pike I use one or more scrolling distortion textures to manipulate the UV's per particle/quad.
open leaf
@meager pelican no, i found in Unity docs that enabling CR is done through RasterState, but the only way to actually set this that i've found is through ScriptableRenderingContext.DrawRenderers
meager pelican
Yeah, but once it's set you should be able to use that semantic to get more info on the pixel (theoretically, I have no idea, haven't tried it).
That's the shader side I though you were asking about.
As to the C# side, IDK. Sorry.
open leaf
I need to overestimate rasterization when rendering mesh triangles onto their "uv-map texture"
meager pelican
I'm exuding optimism here, but sorry I can't help you more. 😉
open leaf
Sure - thanks anyway!
meager pelican
I'd be surprised that it's not enabled per renderer or material. But IDK. Maybe I'll dig. It would be nice if you had a doc reference to save time.
ocean bison
I'm trying to get the light direction of a point light in my shader using the following snippet, but it doesn't seem to be working properly. (I thought this what was my research on this subject suggested, but clearly, I messed it up somehow) float4 LightDir = _WorldSpaceLightPos0; if (LightDir.w == 1) { LightDir.xyz = normalize( i.worldvertpos - LightDir.xyz); LightDir.w = 0; } LightDir.w = 1;//opaque for test return LightDir;//for test
(oh, forgot to metion, works as expected for directional lights, just not point & spot)
meager pelican
Well, output fixed4(LightDir.www, 1) what do you get?
ocean bison
I get the correct color if I'm using a directional light.. and unexpecetedly - NOTHING is displayed when using a point light.
actualy. I didn't do fixed.. hold on
meager pelican
If you're in an opaque queue, the W is probably going to get trashed. That's why I output it into the RGB with a swizzle.
ocean bison
same.. very odd.. I expected somthing to show.. I'm using Tags {"LightMode" = "ForwardBase" "Queue" = "Transparent" "IgnoreProjector" = "True" "RenderType" = "Transparent" }
meager pelican
Did you use the expression I gave you? It should be opaque, white or black.
Not "nothing"
ocean bison
lol.. no- ooops.. I used .xyz, not www- I'll try that.
meager pelican
And you'll want to remove that line before the return that forces it to 1
low lichen
The ForwardBase pass will only pass in information about the directional light, baked lighting, vertex lights and ambient
You need a ForwardAdd pass if you want to draw additional pixel lights
ocean bison
I used.. return fixed4(_WorldSpaceLightPos0.www,1);.. oh!
meager pelican
That's what I thought he was saying.
I was trying to see what happened with your if()
But never mind, MS is correct.
You need proper passes
ocean bison
So do I still need a ForwardBase pass at all?
meager pelican
Yeeah...
low lichen
ForwardAdd adds onto the ForwardBase
ocean bison
hmm.. so If I return (0,0,0,0) from ForwardBase.... then whatever I have in ForwardAdd will be what gets displayed?
low lichen
Why would you want to only use ForwardAdd?
ocean bison
I'm drawing atmpsphere effects- without a light- there is nothing really to draw
works fine for my directional light, but not point light.. all I'm trying to do it make it work with point light also
(I need to replace the directional light in my scene with a point light)
should I do directional light in ForwardBase. THEN the point light in ForwardAdd?
low lichen
That's what Unity's built-in forward rendering expects you to do
But of course, that additive blending for additional lights isn't always what you want
willow pike
@meager pelican thanks! Sorry for the delayed reply. Doesn’t instantiating materials get expensive on performance? I don’t really want to adjust the material manually on each object. Just want the timing of the fire to vary so they don’t look the same. Even if it was randomized that would be fine.
low lichen
@ocean bison Then there isn't really any other option than to pass the additional light information to the shader yourself
willow pike
Could you animate the material based on it position so they vary without creating duplicate materials?
meager pelican
should I do directional light in ForwardBase. THEN the point light in ForwardAdd?
@ocean bison I'd try that.
@willow pike I'd try that.
😉
ocean bison
Just switching to ForwardAdd allows the point light to work as expected- but breaks the directional. So if i need BOTH, I'll need BOTH passes... least thats what it's looking like.
meager pelican
dart, yes. So you don't need per-instance data. You can use Material Property Blocks and not dup materials, but that can have some issues in the new pipeline if you're using certain batchers for some damn reason. But position might just work too. It's all fake anyway. 😉
Agree, Glurth
ocean bison
Thanks guys!!
low lichen
@ocean bison Just be aware that the ForwardAdd pass will be drawn for each light that affects the object, leading to a lot of draw calls with many lights.
This is something that URP does better, it combines the base and add pass into one pass
ocean bison
@low lichen I did see this guy in the docs.. unity_4LightPosX0, unity_4LightPosY0, unity_4LightPosZ0 float4 (ForwardBase pass only) world space positions of first four non-important point lights. perhaps I should use that in ForwardBase instead?
low lichen
That's for vertex lights
Though, I guess there's nothing stopping you from computing them as pixel lights
But you'd have to set them all to non-important/vertex lights
Though I think now unity_LightPosition[4] array is used
ocean bison
is that (non-important) a setting on the light in the scene? (not seeing it)
low lichen
Render Mode
ocean bison
ah, ty!
grand jolt
I need to render a transparent object AFTER (because both the rectangle and circle are buttons, so I need to render the circle "last"). But, I want the circle's transparency to work (so it will effectively replace the pixels on the rectangle (aka delete them). Is there a way to do this?