Game Development Stack Exchange is a question and answer site for professional and independent game developers. It only takes a minute to sign up.
Sign up to join this community
My question is what is written in title. This term "vertex color" appear a lot. I am having hard time to understand this.
For example:
Isn't a vertex simple point on 3D space? If so, how can a point have a color? Or does this "color" mean something else?
There is a "Vertex Color" node in unreal engine's material editor.
Can someone shed light on my question? I couldn't find anything usefull when I googled "what is a vertex color" but the documentations that are too technical for me.
Vertex colors aren't that relevant when you have a fully textured 3d mesh. But it gets interesting when you have an untextured but colored mesh. In that case you assign a color to every vertex. The shader would then color each pixel of a polygon by interpolating between the colors of the three vertices.
Vertex colors can also sometimes be interesting in combination with textures. When you want to use a shading algorithm like Gouraud Shading, you just calculate the light intensity on each vertex, assign the light color to the vertex as a vertex color, and when you render the texture you multiply the color value of each texture pixel with the interpolated colors. This allows you to calculate light sources in the vertex shader instead of the pixel shader. This is usually far faster, because the vertex shader is usually executed far less frequently.
In theory, Vertex Color allows to color a model without having to bother with textures. This is quite common in CAD.
In practice, it's just one more slot you can use to push data into the vertex, which you then use in the vertex shader to do crazy stuff.
"Isn't a vertex simple point on 3D space?"
That's what I used to think, when I first started with 3D. I learned quickly that I was wrong.
You can think of a vertex as a data blob that makes up a corner of a triangle you want to paint on screen (and neither the "corner" part nor the "triangle", "paint", or "screen" parts are always true). Position is one -optional - aspect of that data blob we call Vertex.
There is a lot you can do by attaching additional data to the vertex. The most famous example is that you can add uv coordinates to draw a texture on that triangle. You (usually) can't decide what part of a texture to draw on the triangle if you just have a point - you need texture coordinates too.
Other standars examples of what's in that blob are normals, and tangents. What's important is that all of them are optional, including the position. Vertex Color is just one of these optional inputs that you can use or abuse to create a pretty picture on the screen.
Vertices can have colors, it's a property just like the normal of it.
In 3d enviroments a triangle gets colored based on their vertices' color information. The closest fragments to vertex A get the color of it, the closesr to vertex B get the colornof B, and the color interpolates between the 2.
It works the same way as per-vertex lighting
It's an easy way to add colors to a model without using a texture.
There are solid answers here already - I just want to chime in with one more common way you'll see vertex colours used:
Varying object colours (or other attributes) within a single batch / draw call.
You'll see this used in rendering particle systems, text, and batched (usually atlassed) sprites.
Imagine you have a particle emitter that spawns particles with a random start colour, or that fade to 0 alpha over the particles' lifetime. If we were to use shader uniforms to send this colour data, we'd need to draw each differently-coloured particle in its own draw call, making the particle system much more expensive. Instead, we can bake these colours into the vertex data for the particles, and render the whole batch together in a single call.
Unity appears to use this technique for particle colour variation, as well as for colouring individual characters in a TextMesh or Sprites rendered together in a batch.
I believe you can do something similar with geometry instancing, but I don't know how the two compare in performance. It's possible that today doing this type of instanced tweaking through vertex colours is more of a hold-over to support pre- OpenGL 3.1 / DX9 hardware like some mobile devices, rather than a best practice for modern PCs. I'd welcome comments from folks who've done more performance tuning with batches & instancing to help clear up whether these techniques have superseded vertex colours for this purpose. :)
