Complete Toon Shader
NEW GODOT 4 VERSION AVAILABLE: https://godotshaders.com/shader/complete-cel-shader-for-godot-4/
This is a toon shader with support for almost everything Godot spatial materials can do.
On a basic level, it has:
- Support for multiple lights and tones so you can light your scenes on a more classic fashion;
- Diffuse curve texture for deeper customization like multiple shade bands;
- Toonified specular light blobs;
- Toonified rim light bands;
- 3D outlines. (See the gitlab project, this is not covered on this code)
On a more advanced level, it supports the following Godot capabilities:
- Reflections;
- Emission;
- Normal map;
- Depth map;
- Ambient occlusion map;
- Anisotropy flowchart;
- Subsurface scattering;
- Transmission;
- Refraction.
Read the comments on the code carefully, if you don’t intend on using all of these at the same time you’ll have to remove some lines. You can download the repository and mess around the demo to see things working to understand, or watch the video for some information as well.
Lastly, I want to preface that a most of the code you see here was gathered from around different sources I could find online. I did some modifications to fit them all together, but at its core, this project is a gathering of knowledge in one spot to make it easier to study. Praise everyone you see in the credits on the gitlab repository.
Shader code
shader_type spatial;
render_mode depth_draw_always; // Remove this line if you're not on the transparency pipeline.
uniform vec4 albedo : hint_color = vec4(1.0);
uniform sampler2D texture_albedo : hint_albedo;
// Diffuse curve. This is, in my opinion, what defines a toon shader. A
// photo-realistic shader should have a linear curve starting at 0 and
// finishing at 1, and a toon shader should start at 0 and get to 1 almost
// instantaneously. The length of this transition is its smoothing, working
// the same way as the specular and rim light smoothing uniforms. Another
// common curve is to make it a staircase, which would result in multiple
// shading bands. Check the video for a more detailed explanation.
uniform sampler2D diffuse_curve : hint_white;
// Specular light uniforms. Set specular to zero to turn off the effect.
// The texture map uses the red channel for the specular value, green for amount
// and blue for smoothness.
uniform float specular : hint_range(0,1) = 0.5;
uniform float specular_amount : hint_range(0,1) = 0.5;
uniform float specular_smoothness : hint_range(0,1) = 0.05;
uniform sampler2D texture_specular : hint_white;
// Rim light uniforms. Set rim to zero to turn off the effect.
// The texture map uses the red channel for the rim value, green for rim amount
// and blue for smoothness.
uniform float rim : hint_range(0,1) = 0.5;
uniform float rim_amount : hint_range(0,1) = 0.2;
uniform float rim_smoothness : hint_range(0,1) = 0.05;
uniform sampler2D texture_rim : hint_white;
// Roughness and metallic here are only for reflection purposes if you are
// using SS reflections or want the sky reflected on your material. In most
// cases you don't want to change those, but you can try them. The surface
// texture maps the red channel to roughness and the green channel to metallic.
uniform float metallic : hint_range(0,1) = 0.0;
uniform float roughness : hint_range(0,1) = 1.0;
uniform sampler2D texture_surface : hint_white;
// Emission from base code.
uniform vec4 emission : hint_color = vec4(0.0, 0.0, 0.0, 1.0);
uniform float emission_energy = 1.0;
uniform sampler2D texture_emission : hint_black_albedo;
// Normal map from base code.
uniform float normal_scale : hint_range(-16,16) = 1.0;
uniform sampler2D texture_normal : hint_normal;
// Depth map from base code.
uniform float depth_scale = 0.0;
uniform int depth_min_layers = 8;
uniform int depth_max_layers = 32;
uniform bool flip_tangent = false;
uniform bool flip_binormal = false;
uniform sampler2D texture_depth : hint_black;
// Ambient occlusion from base code.
uniform float ao_light_affect: hint_range(0,1) = 0.0;
uniform sampler2D ao_map : hint_white;
// Anisotropy from base code.
uniform float anisotropy_ratio: hint_range(-1,1) = 0.0;
uniform vec3 anisotropy_direction = vec3(0.0, -1.0, 0.0);
uniform float aniso_map_dir_ratio: hint_range(0,1) = 0.0;
uniform sampler2D anisotropy_flowmap : hint_aniso;
// Subsurface scattering, from base code.
uniform float subsurface_scattering : hint_range(0,1) = 0.0;
uniform sampler2D texture_sss : hint_white;
// Transmission, from base code.
uniform vec4 transmission : hint_color = vec4(0.0, 0.0, 0.0, 1.0);
uniform sampler2D texture_transmission : hint_black;
// Refraction from base code.
uniform float refraction : hint_range(-16,16) = 0.00;
const vec4 refraction_texture_channel = vec4(1.0, 0.0, 0.0, 0.0); // Refraction channel is set to red.
uniform sampler2D texture_refraction;
// UV scale and offset from base code.
uniform vec2 uv_scale = vec2(1,1);
uniform vec2 uv_offset = vec2(0,0);
// Vertex function to deal with UV scale and offset, straight out of base code.
void vertex() {
UV = UV * uv_scale.xy + uv_offset.xy;
}
void fragment() {
vec2 base_uv = UV;
// Depth from base code, with deep parallax enabled. Slightly modified so that
// black are deeps and white are peaks, as well as turning depth flip into bools.
// It is very intense on the GPU, so keep depth scale at zero to turn this off.
if (depth_scale != 0.0) {
vec2 depth_flip = vec2(flip_tangent ? -1.0 : 1.0, flip_binormal ? -1.0 : 1.0);
vec3 view_dir = normalize(normalize(-VERTEX)*mat3(TANGENT*depth_flip.x,-BINORMAL*depth_flip.y,NORMAL));
float num_layers = mix(float(depth_max_layers),float(depth_min_layers), abs(dot(vec3(0.0, 0.0, 1.0), view_dir)));
float layer_depth = 1.0 / num_layers;
float current_layer_depth = 0.0;
vec2 P = view_dir.xy * depth_scale;
vec2 delta = P / num_layers;
vec2 ofs = base_uv;
float depth = 1.0 - textureLod(texture_depth, ofs,0.0).r;
float current_depth = 0.0;
while(current_depth < depth) {
ofs -= delta;
depth = 1.0 - textureLod(texture_depth, ofs,0.0).r;
current_depth += layer_depth;
}
vec2 prev_ofs = ofs + delta;
float after_depth = depth - current_depth;
float before_depth = 1.0 - textureLod(texture_depth, prev_ofs, 0.0).r - current_depth + layer_depth;
float weight = after_depth / (after_depth - before_depth);
ofs = mix(ofs,prev_ofs,weight);
base_uv=ofs;
}
ROUGHNESS = roughness * texture(texture_surface, base_uv).r;
METALLIC = metallic * texture(texture_surface, base_uv).g;
// Normal map, straight out of base code.
NORMALMAP = texture(texture_normal, base_uv).rgb;
NORMALMAP_DEPTH = normal_scale;
// Ambient occlusion, straight out of base code on the red channel.
AO = texture(ao_map, base_uv).r;
AO_LIGHT_AFFECT = ao_light_affect;
// Emission, straight out of base code with additive mode.
EMISSION = (emission.rgb + texture(texture_emission, base_uv).rgb) * emission_energy;
// Subsurface scattering, straight out of base code. I put it on the transparency
// shader because it is a transparency effect, but it actually doesn't work if the
// object is on the transparent pipeline. If you want to use that effect, comment
// out the refraction code and leave the alpha value commented out as well.
SSS_STRENGTH = subsurface_scattering * texture(texture_sss, base_uv).r;
// Transmission, straight out of base code. It works best when on the
// transparent pipeline or with shadows disabled, just make sure it is
// not showing its own shadow on its back.
TRANSMISSION = transmission.rgb + texture(texture_transmission, base_uv).rgb;
// Refraction, slightly modified. It doesn't actually uses the alpha value, instead
// it reads from the screen texture in order to distort the image to make refraction.
// This means nothing else that reads from it will appear through it, and things on the
// alpha channel also have troubles showing behind them. If you want to disable refraction
// and just use the alpha channel, just take out all of the following uncommented lines
// and uncomment the commented ones.
vec3 normal = normalize(mix(NORMAL,TANGENT*NORMALMAP.x + BINORMAL*NORMALMAP.y + NORMAL*NORMALMAP.z,NORMALMAP_DEPTH));
vec2 ref_ofs = SCREEN_UV - normal.xy * dot(texture(texture_refraction, base_uv), refraction_texture_channel) * refraction;
float ref_amount = 1.0 - albedo.a * texture(texture_albedo, UV).a;
EMISSION += textureLod(SCREEN_TEXTURE, ref_ofs, 0.0).rgb * ref_amount;
ALBEDO = (1.0 - ref_amount) * albedo.rgb * texture(texture_albedo, UV).rgb;
ALPHA = 1.0;
// ALBEDO = albedo.rgb * texture(texture_albedo, UV).rgb;
// ALPHA = albedo.a * texture(texture_albedo, UV).a;
}
const float PI = 3.14159265358979323846;
void light() {
// Let's start by incorporating specular and rim textures. Pay attention to
// the channels and what each value does.
float spec_value = specular * texture(texture_specular, UV).r;
float spec_gloss = pow(2.0, 8.0 * (1.0 - specular_amount * texture(texture_specular, UV).g));
float spec_smooth = specular_smoothness * texture(texture_specular, UV).b;
float rim_value = rim * texture(texture_rim, UV).r;
float rim_width = rim_amount * texture(texture_rim, UV).g;
float rim_smooth = rim_smoothness * texture(texture_rim, UV).b;
// Diffuse part. We take the dot product between light and normal, multiply it by attenuation
// and apply it to the diffuse curve. This means the diffuse curve gets to do the dot product
// smoothing, set multiple light bands each with its own tone and smoothing, etc etc. I reccomend
// using the gradient tool to make a curve, since it gives you control of each point's position
// and color with precision. The curve tool works too, it gives you control of different
// interpolation methods but you have less control over each point's exact position and value.
vec3 litness = texture(diffuse_curve, vec2(dot(LIGHT, NORMAL), 0.0)).r * ATTENUATION;
DIFFUSE_LIGHT += ALBEDO * LIGHT_COLOR * (litness + TRANSMISSION * (ATTENUATION - litness));
// Specular part. We use the Blinn-Phong specular calculations with a smoothstep
// function to toonify. Mess with the specular uniforms to see what each one does.
// It also deals with anisotropy. If you want to remove anisotropy calculations,
// remove flowchart, aniso_dir, aniso and replace spec by dot(NORMAL, half).
vec3 half = normalize(VIEW + LIGHT);
vec3 flowchart = (texture(anisotropy_flowmap, UV).rgb * 2.0 - 1.0);
vec3 aniso_dir = mix(normalize(anisotropy_direction), flowchart, aniso_map_dir_ratio);
float aniso = max(0, sin(dot(normalize(NORMAL + aniso_dir), half) * PI));
float spec = mix(dot(NORMAL, half), aniso, anisotropy_ratio * texture(anisotropy_flowmap, UV).a);
float spec_intensity = pow(spec, spec_gloss * spec_gloss);
spec_intensity = smoothstep(0.05, 0.05 + spec_smooth, spec_intensity);
SPECULAR_LIGHT += LIGHT_COLOR * spec_value * spec_intensity * litness;
// Rim part. We use the view and normal vectors only to find out if we're looking
// at a pixel from the edge of the object or not. We add the final value to specular
// light values so that Godot treats it as specular.
float rim_dot = 1.0 - dot(NORMAL, VIEW);
float rim_threshold = pow((1.0 - rim_width), dot(LIGHT, NORMAL));
float rim_intensity = smoothstep(rim_threshold - rim_smooth/2.0, rim_threshold + rim_smooth/2.0, rim_dot);
SPECULAR_LIGHT += LIGHT_COLOR * rim_value * rim_intensity * litness;
}
Hey, this is a great shader!!! Thank you for sharing this!
Thank you , love ya
Project no longer seems run/compile under 3.5.1 ?
Is it possible to update this shader to Godot 4? I managed to make it work there but it totally looks bugged in comporison to 3.5.1. Especially the diffuse curve does not work as intented.
Works fine on 4rc. I did have to change the keyword ‘hint_color’ to ‘source_color’ for anyone else attempting to use it.
This is lovely! Would love to see a version for godot 4! Can’t seem to port it to 4 quite right :/
i’m completely layman about shaders performance: if i remove from code the things i won’t use in my project, like reflections, AO, etc, will it perform better?
Yes. This is how the demo works, there are multiple versions of the shader, each version containing different blocks fo code from this whole. This is also most of the biggest changes on Godot 4 with preprocessor, which allows us to quickly add/remove blocks of code. To try to explain a little bit why this is important, know that GPUs are built to do thousands of processings at the same time. The vertex() function runs once for each vertex, every run happening at the same time in parallel. After vertex() finishes, fragment() runs per pixel, and then light() runs per pixel per light source. You can imagine how one additional operation can make a difference. In short, you wanna simplify everything and make as few operations as possible. You also want to stay away from ifs and for loops most of the time.