Writing shaders that interact with lighting is complex. There are different light types, different shadow options, different rendering paths (forward and deferred rendering), and the shader should somehow handle all that complexity.
Surface Shaders in Unity is a code generation approach that makes it much easier to write lit shaders than using low level vertex/pixel shader programs. Note that there are no custom languages, magic or ninjas involved in Surface Shaders; it just generates all the repetitive code that would have to be written by hand. You still write shader code in Cg / HLSL.
For some examples, take a look at Surface Shader Examples and Surface Shader Custom Lighting Examples.
You define a “surface function” that takes any UVs or data you need as input, and fills in output structure SurfaceOutput. SurfaceOutput basically describes properties of the surface (it’s albedo color, normal, emission, specularity etc.). You write this code in Cg / HLSL.
Surface Shader compiler then figures out what inputs are needed, what outputs are filled and so on, and generates actual vertex&pixel shaders, as well as rendering passes to handle forward and deferred rendering.
Standard output structure of surface shaders is this:
struct SurfaceOutput
{
fixed3 Albedo; // diffuse color
fixed3 Normal; // tangent space normal, if written
fixed3 Emission;
half Specular; // specular power in 0..1 range
fixed Gloss; // specular intensity
fixed Alpha; // alpha for transparencies
};
In Unity 5, surface shaders can also use physically based lighting models. Built-in Standard and StandardSpecular lighting models (see below) use these output structures respectively:
struct SurfaceOutputStandard
{
fixed3 Albedo; // base (diffuse or specular) color
fixed3 Normal; // tangent space normal, if written
half3 Emission;
half Metallic; // 0=non-metal, 1=metal
half Smoothness; // 0=rough, 1=smooth
half Occlusion; // occlusion (default 1)
fixed Alpha; // alpha for transparencies
};
struct SurfaceOutputStandardSpecular
{
fixed3 Albedo; // diffuse color
fixed3 Specular; // specular color
fixed3 Normal; // tangent space normal, if written
half3 Emission;
half Smoothness; // 0=rough, 1=smooth
half Occlusion; // occlusion (default 1)
fixed Alpha; // alpha for transparencies
};
See Surface Shader Examples, Surface Shader Custom Lighting Examples and Surface Shader Tessellation pages.
Surface shader is placed inside CGPROGRAM..ENDCG block, just like any other shader. The differences are:
#pragma surface ... directive to indicate it’s a surface shader.The #pragma surface directive is:
#pragma surface surfaceFunction lightModel [optionalparams]
void surf (Input IN, inout SurfaceOutput o), where Input is a structure you have defined. Input should contain any texture coordinates and extra automatic variables needed by surface function.Standard and StandardSpecular, as well as simple non-physically based Lambert (diffuse) and BlinnPhong (specular). See Custom Lighting Models page for how to write your own.
Standard lighting model uses SurfaceOutputStandard output struct, and matches the Standard (metallic workflow) shader in Unity.StandardSpecular lighting model uses SurfaceOutputStandardSpecular output struct, and matches the Standard (specular setup) shader in Unity.Lambert and BlinnPhong lighting models are not physically based (coming from Unity 4.x), but the shaders using them can be faster to render on low-end hardware.
Transparency and alpha testing is controlled by alpha and alphatest directives. Transparency
can typically be of two kinds: traditional alpha blending (used for fading objects out) or more
physically plausible “premultiplied blending” (which allows semitransparent surfaces to retain proper
specular reflections). Enabling semitransparency makes the generated surface shader code contain
blending commands; whereas enabling alpha cutout will do a fragment discard in the
generated pixel shader, based on the given variable.
alpha or alpha:auto - Will pick fade-transparency (same as alpha:fade) for simple lighting functions, and premultiplied transparency (same as alpha:premul) for physically based lighting functions.alpha:fade - Enable traditional fade-transparency.alpha:premul - Enable premultiplied alpha transparency.alphatest:VariableName - Enable alpha cutout transparency. Cutoff value is in a float variable with VariableName. You’ll likely also want to use addshadow directive to generate proper shadow caster pass.keepalpha - By default opaque surface shaders write 1.0 (white) into alpha channel, no matter what’s output in the Alpha of output struct or what’s returned by the lighting function. Using this option allows keeping lighting function’s alpha value even for opaque surface shaders.decal:add - Additive decal shader (e.g. terrain AddPass). This is meant for objects that lie atop of other surfaces, and use additive blending.decal:blend - Semitransparent decal shader. This is meant for objects that lie atop of other surfaces, and use alpha blending.Custom modifier functions can be used to alter or compute incoming vertex data, or to alter final computed fragment color.
vertex:VertexFunction - Custom vertex modification function. This function is invoked at start of generated vertex shader, and can modify or compute per-vertex data See Surface Shader Examples.finalcolor:ColorFunction - Custom final color modification function. See Surface Shader Examples.Shadows and Tessellation - additional directives can be given to control how shadows and tessellation is handled.
addshadow - Generate a shadow caster pass. Commonly used with custom vertex modification, so that shadow casting also gets any procedural vertex animation. Often shaders don’t need any special shadows handling, as they can just use shadow caster pass from their fallback.fullforwardshadows - Support all light shadow types in Forward rendering path. By default shaders only support shadows from one directional light in forward rendering (to save on internal shader variant count). If you need point or spot light shadows in forward rendering, use this directive.tessellate:TessFunction - use DX11 GPU tessellation; the function computes tessellation factors. See Surface Shader Tessellation for details.Code generation options - by default generated surface shader code tries to handle all possible lighting/shadowing/lightmap scenarios. However in some cases you know you won’t need some of them, and it is possible to adjust generated code to skip them. This can result in smaller shaders that are faster to load.
exclude_path:deferred, exclude_path:forward, exclude_path:prepass - Do not generate passes for given rendering path (Deferred Shading, Forward and Legacy Deferred respectively).noshadow - Disables all shadow receiving support in this shader.noambient - Do not apply any ambient lighting or light probes.novertexlights - Do not apply any light probes or per-vertex lights in Forward rendering.nolightmap - Disables all lightmapping support in this shader.nodynlightmap - Disables runtime dynamic global illumination support in this shader.nodirlightmap - Disables directional lightmaps support in this shader.nofog - Disables all built-in Fog support.nometa - Does not generate a “meta” pass (that’s used by lightmapping & dynamic global illumination to extract surface information).noforwardadd - Disables Forward rendering additive pass. This makes the shader support one full directional light, with all other lights computed per-vertex/SH. Makes shaders smaller as well.Miscellaneous options
softvegetation - Makes the surface shader only be rendered when Soft Vegetation is on.interpolateview - Compute view direction in the vertex shader and interpolate it; instead of computing it in the pixel shader. This can make the pixel shader faster, but uses up one more texture interpolator.halfasview - Pass half-direction vector into the lighting function instead of view-direction. Half-direction will be computed and normalized per vertex. This is faster, but not entirely correct.approxview - Removed in Unity 5.0. Use interpolateview instead.dualforward - Use dual lightmaps in forward rendering path.To see what exactly is different from using different options above, it can be helpful to use “Show Generated Code” button in the Shader Inspector.
The input structure Input generally has any texture coordinates needed by the shader. Texture coordinates must be named “uv” followed by texture name (or start it with “uv2” to use second texture coordinate set).
Additional values that can be put into Input structure:
float3 viewDir - will contain view direction, for computing Parallax effects, rim lighting etc.float4 with COLOR semantic - will contain interpolated per-vertex color.float4 screenPos - will contain screen space position for reflection or screenspace effects.float3 worldPos - will contain world space position.float3 worldRefl - will contain world reflection vector if surface shader does not write to o.Normal. See Reflect-Diffuse shader for example.float3 worldNormal - will contain world normal vector if surface shader does not write to o.Normal.float3 worldRefl; INTERNAL_DATA - will contain world reflection vector if surface shader writes to o.Normal. To get the reflection vector based on per-pixel normal map, use WorldReflectionVector (IN, o.Normal). See Reflect-Bumped shader for example.float3 worldNormal; INTERNAL_DATA - will contain world normal vector if surface shader writes to o.Normal. To get the normal vector based on per-pixel normal map, use WorldNormalVector (IN, o.Normal).Currently some parts of surface shader compilation pipeline do not understand DirectX 11-specific HLSL syntax, so if you’re using HLSL features like StructuredBuffers, RWTextures and other non-DX9 syntax, you have to wrap it into a DX11-only preprocessor macro. See Platform Specific Differences page for details.