nooshu - Matt Hobbs' Web Development Blog

Kneeling on the shoulders of giants

Category: Three.js

A visualisation experiment using SoundCloud, Backbone.js, three.js and node.js

Client work, as always, has been busy over the past month or so; so it’s always refreshing to have a play with something a little different. It’s even better when you can actually combine the two! Visual Jazz Isobar decided to have a little code competition open to all the developers in the company. The brief was very simple: build something cool using an external API (bonus points for multiple API’s). Very open brief don’t you think? You only have to have a look at the list of API’s available on programmable web to see that.

As I’ve recently been looking at Backbone.js I decided to put what I’ve learned good use. The experiment evolved over a couple of weeks. I originally planned to use d3.js and last.fm but eventually settled on three.js and SoundCloud. My final submission looked like this:

Visualisation using Backbone.js, Node.js and Three.js

My little experiment using the SoundCloud API and three.js (with a sprinkling of node.js)

To be honest I could have built the experiment easily without using Backbone, but where is the challenge in that! I love the whole Model, View, Router set up. You do seem to be writing a lot of code a first for very little functionality, but once you are over the initial set up you can add features very quickly. I’ll definitely be using it again for future projects.

The visualisation itself makes use of the Web Audio API. The API is still being developed and standardized, because of that it only runs in Chrome at the moment. Three.js itself runs fine in Firefox, but no audio visualisation.

During development I stumbled across an issue with Chrome and the way the audio buffer works; for the data to load, the music files need to be on the same server as the application. This was a bit of a problem as I was using the SoundCloud API for music streaming. That’s where node.js and Nodester came to the rescue. With a little bit of node.js magic it’s fairly simple to create a proxy server and ‘bounce’ the stream from SoundCloud via the server to the visualisation. A big downside to this method is the music stream (mp3) has to be fully downloaded before the visualisation starts. For this reason only tracks under 6 minutes are returned as search results from SoundCloud. It’s not ideal but it work!

When the competition results came in I was placed second; not first but hey I learnt a lot in the process! You can take a look at my little experiment here.

Using the Fullscreen and Pointer Lock API’s

This post is a couple of months late due to real life getting in the way, but I finally got around to experimenting with the Fullscreen and Pointer Lock API (formally known as the Mouse Lock API). Both are still in the experimental stage, Pointer Lock especially. To get it to work you will have to be running the latest Canary build of Chrome (currently 19.0.1067.0) and enable it via about:flags. I’ve had no luck getting Pointer Lock working in the latest version of Firefox and Aurora, even though most documentation on the API is written by Mozilla. Maybe there’s a setting I’m missing?

So what exactly do the Fullscreen and Pointer Lock API’s do you may ask. Well Fullscreen is obvious; it allows you to view any web content in fullscreen mode, removing the browser UI for distraction free reading. Pointer Lock gives a developer access to raw mouse movement data, not just the absolute position of the cursor. You are no longer limited to the size of the browser window, so you can scroll continuously in one direction without having to stop. Where both of these API’s come in very handy is the development of HTML5 games. It’s now possible to create a games like Quake 3 directly in the browser with usable controls… impressive!

To have a play with both API’s I put together a little experiment using three.js and a panoramic image.

Fullscreen, Pointer Lock and three.js

What a beautiful day in Sydney, just round the corner from where I live.

I started by building on an excellent blog post by Paul Lewis called “Create your own environment maps”. Using Microsoft’s Photosynth App and Blender it’s dead simple to create a panoramic image. After adapting Paul’s code to my needs, I added the Fullscreen and Pointer lock API’s. To enable Fullscreen mode press enter. If you have a compatible browser, the browser UI will disappear and you will enter fullscreen mode. If Pointer Lock is available, a prompt will appear asking for permission to hide your cursor. You will now be able to scroll left and right continuously without having to stop, if Pointer Lock isn’t available the demo will drop back to standard mouse co-ordinates.

A quick tip for working with both of these API’s: use a shim like that made by Brandon Jones, it will save you a lot of cross browser pain! Browser vendors are still changing the way both of these API’s are implemented, so until this stabilises, using a shim is the way to go. Big thank you to Paul and Brandon for their excellent contributions.

Plotting the Lorenz Attractor using three.js

It’s been a while since I dabbled with three.js. There have been a few API changes, so it’s slightly different from what I remember. That being said it’s still as fun to use, and there are even a few API docs available (not sure if they are fully up to date though)! I happened to be reading an article on Chaos Theory and the “butterfly effect” which mentioned a lovely set of equations; the Lorenz Equations, also known as the Lorenz oscillator. So I decided to investigate further. Here they are in all their glory:

\begin{aligned}\dot{x} & = \sigma(y-x) \\ \dot{y} & = \rho x - y - xz \\ \dot{z} & = -\beta z + xy \end{aligned}

So what exactly is the Lorenz oscillator? Well to quote from Wikipedia “The Lorenz oscillator is a 3-dimensional dynamical system that exhibits chaotic flow, noted for its figure eight shape.”. What this really boils down to is how even very small change in the equations initial conditions (x, y and z) causes large changes over time, in a complex and non-repeating pattern. There’s a famous title of a talk given in 1972 by Philip Merilees:

Does the flap of a butterfly’s wings in Brazil set off a tornado in Texas?

The butterfly flapping its wings introduces slight changes to the initial conditions, creating a chain of events that lead to a tornado in Texas. The Lorenz butterfly is a graphical way of showing these changes over time on a much smaller scale.

I’ve created a demo that allows you to change variables related to the Lorenz butterfly and observe the effect it has on the system.

Example of my Lorenz attractor demo in action

Adjust the demo variables to see how the Lorenz butterfly changes.

There are a few variables you can play to change how the Lorenz attractor is rendered:

  • rhoValue: Called the Rayleigh number. Rho of value 28 shows chaotic behavior. Other values show periodic orbits.
  • showAxis: Allows you to turn the x, y, z (red, green, blue) axis on for reference.
  • randomStart: Picks a random initial conditions, rather than 0, 0, 0.
  • totalTime: The total time (or number of iterations performed) when generating the graph. Higher numbers equals more lines and increases the render time.
  • timeIncriment: The accuracy of the graph depends on the number of points plotted. A lower number creates a smoother graph but increases render time.
  • colorModifier: Simply changes the colo(u)r. Pointless but pretty!

While rendering a larger number of points, I noticed the browser lock-up for a second or so. This was because the points were being calculated on the browsers UI thread. Luckily modern browsers support thread like message passing in the form of the Web Workers API. By offloading the point calculation to its own thread (lorenzPoints.js), I was able to stop the nasty UI freeze. Unfortunately you are fairly limited with what you can pass back and forth between workers. It was possible to generate the three.js geometry in the worker thread thanks to the importScripts() method. I tried passing the three.js geometry back to the browser but all the methods were stripped out, leading to lots of errors (JSON issues). In the end I dropped back to calculating the points in the worker, pushing them into an array and generating the geometry on the main browser thread.

It was good fun dabbling with three.js, Web Workers and dat.gui again. Now on to my next project… View the demo here.

Antimatter 3D Graph Plotter and a little animation

Last year I stumbled upon this blog post by @antimatter15 who added a little three line modification to mr.doobs particle floor demo and created a graph plotter in three.js. Unfortunately as three.js has developed and evolved the API has changed; in doing so it broke the graphing demo. I left a message asking if there was a fix on Antimatters blog but got no response. So I decided to update the code and fix it for the latest version of three.js; once updated for the latest API it ran as expected.

3D graph plotter using three.js

Antimatter15 graph plotter updated for latest version of three.js.

You pass the equation you wish to plot via a URL parameter and out the other pops a pretty 3D graph. Here are a couple of examples taken from the original blog post:

In updating the code I thought I’d have a little play myself and see if I could add a bit of motion to the graphs just for the fun of it. I’ve created a version of my own which you can view here.

Graphing multiple for formula using JavaScript and three.js

Modified version allows you to add more than one formula to plot.

In my version you no longer pass the formula you wish to plot via a URL parameter, you add it via the panel on the right hand side. You can add and remove as many equations as you wish and the demo will cycle between them. To do so update the right hand panel then click the “Plot new formula(e)” button. The equation that is currently being plotted is highlighted in grey and available mathematical functions listed below.

With a little bit of trial end error you can create some interesting effects. There may not be much practical use for the experiment but I enjoyed putting it together with a little help from mr.doob and antimatter15 :)

Debug Axes in three.js

Update: The code I had previously stopped working with the latest three.js API. I’ve updated and improved the code for the latest API.

I’ve been working on a couple of small projects using three.js recently; An issue I always seem to have when working with 3D is knowing the current orientation. To help me out I’ve put together a small snippet of JavaScript.

Creating a set of axis in three.js

{x: red, y: green, z: blue} It's not much to look at but the axes can be very useful.

The point where the lines cross is 0,0,0 as you’d expect, you can adjust the length of the line by changing the axisLength variable. I’ve uploaded a small demo so you can see the code used or copy / paste below:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
var debugaxis = function(axisLength){
    //Shorten the vertex function
    function v(x,y,z){
            return new THREE.Vertex(new THREE.Vector3(x,y,z));
    }
   
    //Create axis (point1, point2, colour)
    function createAxis(p1, p2, color){
            var line, lineGeometry = new THREE.Geometry(),
            lineMat = new THREE.LineBasicMaterial({color: color, lineWidth: 1});
            lineGeometry.vertices.push(p1, p2);
            line = new THREE.Line(lineGeometry, lineMat);
            scene.add(line);
    }
   
    createAxis(v(-axisLength, 0, 0), v(axisLength, 0, 0), 0xFF0000);
    createAxis(v(0, -axisLength, 0), v(0, axisLength, 0), 0x00FF00);
    createAxis(v(0, 0, -axisLength), v(0, 0, axisLength), 0x0000FF);
};

//To use enter the axis length
debugaxis(100);

It’s not much to look at, but it can certainly help you get your bearing when starting a new project (or when things aren’t working!). Hopefully you find it useful.

Currently on page 1 of 212