How do I use a Heightmap in Shader Graph?

Hello,
I would like to add a Height Map to my Shader i created with Shader Graph, but it is missing on the Master. How do I add a Height Map?

 

The original Standard shader's height map parameter is used to drive a version of Parallax Offset Mapping, sometimes just called Offset Mapping, or in Unity's code referred to as Parallax Bump Mapping. This is a simple approximation that basically just smudges the UVs a little based on a height map.

Here's the code Unity uses:

Copied from UnityCG.cginc

Code (CSharp):

1. // Calculates UV offset for parallax bump mapping
2. inline float2 ParallaxOffset( half h, half height, half3 viewDir )
3. {
4.     h = h * height - height/2.0;
5.     float3 v = normalize(viewDir);
6.     v.z += 0.42;
7.     return h * (v.xy / v.z);
8. }

You can add this yourself by building a node graph to do the same. the h is the greyscale value from the height map texture, the height is the scale factor (which should be very, very small; Standard uses a range of 0.005 to 0.08), and the viewDir is the tangent space view direction. The output value should be added to the texture UV.

The HD pipeline's Lit shaders also have a height map, but it uses it for tessellation displacement (which isn't supported by Shader Graph yet), or for Parallax Occlusion Mapping, which is a much more accurate technique that raymarches through the height field to find an appropriate UV, and is much more expensive than Parallax Offset Mapping. It also requires the use of dynamic loops which node graphs aren't great at handling. And by not great I mean cannot be handled at all without a custom node.

If replicating the existing Standard Shader's implementation is good enough, the above code should let you recreate it in your own node graph, or let you create a subgraph you can reuse elsewhere. If you want something like what the HD Lit shaders have you'll have to wait for Unity to add support or you can request a Parallax Occlusion Mapping node.

 

Whole set of nodes for Shader graph
http://u3d.as/1gkf
- SSS Color,
- SSS gradient,
- Parallax mapping,
- Parallax Mapping iterrations,
- Parallax Occlusion Mapping,
- PerturbedNormals (World normal с учетом NormalMap),
- CreateNormalChannel (для каждого канала текстуры),
- CreateNormalFloat (для любого значения float, можно делать нормали из процедурных шумов),
- LightDirection,
- ProceduralRock (процедурный шум),
- Range ,
- TilingShape .

 

I’m a bit confused how to implement a height map in the shader graph, can some one please please provide an example

Thanks so much

 

Hello I would like to know more as well, do you have a documentation somewhere ? I can't seem to find it or reproduce it

 

Hello, instead of documentation there are ready examples, shaders are in the folder "Shaders", you need Parallax.

 

It seems like High map is not supported. Been looking for tessellation without luck.

 

Here is Sub Graph version of this code:

 

This is a great literal conversion - but if the master node doesn't have a parallax input, then where would you plug the output of this sub-graph in your main shader graph?

How would one go about adding some sort of a height field to the LWRP's Master node? Is this just an addition that Unity would be responsible for? The HDRP master lit node doesn't seem to have straight forward parallax support either.

 

As bgolus said, parallax mapping is just a special way of distorting uvs of a texture sample. If you want to add the illusion of height to your shader, you just need to plug the return of the subgraph above into the UVs of your texture sampler.

 

Might be a dumb question, but where could i plug a heightmap texture into this shadergraph?

 

The graph @V0odo0 showed is a Sub Graph, which means it becomes a node in another graph that you would pass the output of other nodes into. To use a texture, add it to the Shader Graph's properties (via the blackboard), then add that texture as a property node along with a Sample Texture 2D node. You'd then pipe the output of that into the Height input of the Sub Shader node.

Alternatively you could just put all of those nodes into your Shader Graph shader directly and replace the Height node with the Sample Texture 2D node.

 

I can't manage to get this shader to work properly,
I put it on a quad and used 2 checker textures just to test, everything works fine until I either rotate the quad or I rotate the UVs, then it becomes a mess. I am using the node setup as translate from the code by VOodoO
How does this apply to local space? I am smashing my head the whole day trying to understand how to get UVs rotation but I bet I'm working in the wrong direction.
Any clues?

 

I bought the asset package from Andrey_Graphics and found the solution, I am still not fit enough to read shader code.
There the node FixedViewDirectionTangent did the trick, I converted it to node tree from code snippet and now I can go on developing my lo-fi parallax.

PS: Andrey, please do some better naming convention on your package, it's really nice but it's a mess

 

@Pivetta I believe you're missing a "height" parameter (the one @bgolus said usually goes from .005 to .08). Use that value to multiply and divide in the top left corner, instead of the checkerboard.

If you do want to use the checkerboard for your height map, plug it into the 'A' slot in the multiply node instead of the Property + Add .5 nodes (I think those just allow you to set a manual offset for the entire texture). @binaristudios you can plug any texture + a texture sampler into that top left slot and it will act as a height map (see attached).

 

Since this is the first thread that appears when searching for Heightmap Shader Graph i'll post my HDRP node setup here, this is Height Map with optional Parallax Occlusion Mapping and Normal Map (with option to generate normal from height, although this tends to be less quality), the UV input on the far left is just the UV Channel the user has selected.

Remember to click the gear icon on the master node and activate the Depth Offset boolean so you can feed in your Parallax Offset to the master node.

This will automatically Parallax the incoming UVs if you have a Parallax Bool enabled. Simply use the output of the Parallax Bool Check on the left side and feed it into all your UV inputs across your maps.

This currently works in HDRP 2020.1.9f1

 

Hey guys, can someone see if they can get this thing working

https://docs.unity3d.com/Packages/com.unity.shadergraph@10.2/manual/Parallax-Mapping-Node.html

 

I have this working nicely but I am applying my adjusted UVs from height map to my normal and other maps. I wanted to pack the height in with something else, i.e. normal, but seems I need the height first to get all the other components at the adjusted UVs. Am I doing something wrong?

 

You can't pack it as your main material textures (albedo, normal, etc) depend on the UVs that have been changed via the parallax equation.

 

Thanks, yeah. I came to this conclusion pretty much, also posted on redit. Basically I can pack and it will save some texture space but I will need to read once to get the height and then read again to get the adjusted normal. Maybe I will get some cache hits occasionally. I can also LOD the surfaces so parallax only comes into play for nearby objects.

 

Yeah it would be interesting to see because yes you will be getting more cache hits, but at the same time, you are also sampling more data than required, as you could store it in a single channel texture instead. I suspect at grazing angles, unpacked will be faster, and at straight-on angles, packed might be faster.

 

Yeah on face value packing it has no huge advantage, might even be slower oddly. I might look at the single channel but I suspect those don't perform as well.

I do see some shimmering artefacts with this. Especially as I get a few units out from the view position in VR and at more acute angles. Anything you can recommend?

I have to say that normal maps in VR do not work well for large scale objects. Parallax Height Mapping looks great but the shimmering is a bit of a deal breaker - probably there are more expensive algorithms I could try.

For smaller objects, for third person view puzzlers and platformers, I think normal maps look good, but for large objects, up close, not really adding a huge amount, it has to be said.