Getting Started with Three.js



I think WebGL is an awesome idea; you can use accelerated 3D graphics within a browser. This is totally game changing and opens the gate to games serves directly to your browser.

Working with OpenGL primitives is somewhat annoying. (One of the reason why I build wrappers for C++.) But for javascript there are already quite usable libraries, one of these is Three.js. Today I will show the relative easy with which you can start to use WebGL. (Easy to learn hard to master.)

I will show you how implement the following:

full screen

The basic template I will be using is the following.

      body { margin: 0; }
      canvas { width: 100%; height: 100% }
    <script src="js/three.min.js"></script>    
        <!-- code -->

This is a HTML skeleton and it does little, except being a valid HTML file. As a start we will just make the rendered to area the entire usable area of the screen. This is enforced by the CSS style in the header.

The actual interesting bits are the script elements. Firstly we need to actually load Three.js. The next script element we will put the actual setup and rendering code. (The following JavaScript code goes there.)

The core of Three.js is it's WebGL renderer and that is the first thing that we will setup.

var renderer = new THREE.WebGLRenderer();
renderer.setSize(window.innerWidth, window.innerHeight);

This is a common paradigm, it create a renderer without a canvas to render to. The renderer will then create the canvas and put it into domElement. To ensure that the canvas is the same size as the window's client area the renderer is instructed to size the canvas appropriately. Finally the actual canvas is then attached to the document's body.

This will only give you a black screen, now we will create somehting to render.

Three.js uses the concept of a scene to structure the thing to render. This is basically a scene graph, although a simple one.

var scene = new THREE.Scene();

When rendering something, you need something to tell the renderer from where to render, this dome by the camera.

var camera = new THREE.PerspectiveCamera(75, window.innerWidth/window.innerHeight, 0.1, 1000);
camera.position.z = 3;

In our case we are using a perspective camera; this is the most common case when rendering realistic scenes. The camera is constructed with the field of view, aspect ratio, the near plane and the far plane. If you want to know how perspective cameras work in OpenGL, here is good article on the subject. The camera is moved up from the centrer, since our subject will be there.

Speaking of subject, here is a cube to look at.

var geometry = new THREE.BoxGeometry(1, 1, 1);
var material = new THREE.MeshPhongMaterial({ color: 0x1C4A8C });
var cube = new THREE.Mesh(geometry, material);

In Three.js the mesh is the primary visible object. The mesh is built up of two things, a geometry that defines it's shape and a material that defines the visual properties of the object. There are multiple types of geometries and materials but for our purposes we will use a box and the almighty phong material.

But since the phong material is fully lit, we also need alight to actually illuminate the scene. For these purposes we will use a simple directional light.

var light = new THREE.DirectionalLight(0xffffff, 0.55);
light.position.set(0, 0, 1);

The light is constructed with the color and intensity. This is useful, since it means you don't have to encode the intensity in the color. Directional lights are odd in Three.js, their position defines the direction of the light. That is the direction of light is the vector form the position to the origin. Here the light shines directly in the same direction as the camera is looking onto the scene.

But so far nothing actually is happening, but here is the render loop:

var render = function () {
  requestAnimationFrame( render );

  cube.rotation.x += 0.01;
  cube.rotation.y += 0.01;

  renderer.render(scene, camera);


The render loop uses the requestAnimationFrame function to ensure continues rendering without taxing the CPU unnecessary. The function also rotates the cube, to show that this actually is dynamic. But the really important bit is the invocation of the actual render function of the renderer.

You can see the full code on github and the entire thing in action.

There are a few remaining issues, such as resizing and attaching to an existing canvas, but that I will address in a further article.