Last week I received my Oculus Rift DK2 headset which I ordered back in August. I had my eye on the Oculus DK1, the first version of their headset, but I was too busy getting Antigen out of the door to let myself be distracted by other projects. Now Antigen is out of the way, I can let myself experiment a little while I figure out what my next big project will be.

It comes in a padded, reusable cardboard box, which is fine for storage but if I was going to take it anywhere I would probably get a Pelican camera case or something similar, and cut some custom foam inserts for it, like this Oculus user has done.

The headset is surprisingly light, and connects to your computer with a sturdy cable that runs over the top of your head along one of the headset’s straps. I imagine this was done for weight distribution and to help keep the cable out of the way when you move around.

The cable connects to your computer via 1 USB and 1 HDMI port. There is a motion tracking IR camera that clips to your monitor like a webcam. This uses 1 USB port. There’s also an optional power supply if you want to use the USB port built into the headset. One device that uses this is the Leap Motion, which I’d like to get my hands on sooner or later…

The Rift also comes with two sets of detachable lenses; a longer lens that is already installed in the headset, and a shorter lens intended for nearsighted users.

Setup was easy. I downloaded the drivers from the Oculus website, installed them and as instructed, updated the headset firmware using a simple and straightforward preferences utility. This tool also lets you create user profiles for the Rift, containing your height and other details.

The Rift appears to your computer as a second monitor. You can view anything you drag on there, but it will look like crap unless it is rendered stereoscopically and positioned correctly. There is a DirectX style “Direct Mode” that lets applications talk directly to the Rift, but it doesn’t work on the Mac yet. (The Mac is a bit of a second-class citizen for the Rift at the moment. Unfortunately, most of the coolest demos are Windows only too.)

At this point I could hardly contain my urge to jack in to the matrix and become the console cowboy I’d always dreamed of being, so I hit “Show Demo” and braced myself.

I found myself sitting at a desk in cyberspace, somewhere on an infinite Tron-esque plain. The first thing that struck me was the stereoscopy, the 3d-ness of the lamp and the tower of playing cards on my desk. I really felt I could reach out and touch them.

The resolution is good at 960×1080 for each eye, but I don’t think it’s high enough for consumer use yet. I suspect that in order to approximate the usability and experience of a desktop monitor, VR displays will require a much higher relative pixel density. There is also a kind of “screen door effect” where you can see the pixel grid. Most of the time this isn’t too bad, but it is noticeable when you try and view details that are further away, such as the faces of game characters.

Text is readable as long as it is large and in the middle of the display. It’s a bit like looking at text on a C64 with a bad TV set. Towards the edges of the display it gets blurry and there is some chromatic aberration. You need to look at text directly in order to read it properly, which will have ramifications for UI design, particularly games where information is usually positioned at the edge of the screen.

All the same, these aren’t really criticisms, they are just statements about where the developers kit is right now. The DK2 is a prototype that a deranged experimenter committed person works with and adapts himself to, rather than a consumer product. I’ve not had a chance to try the new Crescent Bay prototype, but reports suggest that it eliminates the screen door effect and that we can expect the consumer version to be a real advance on what is currently available.

There’s a nice demo included with the SDK where you can walk around an Italian country house and its surrounding grounds. The house is on a cliff over the sea. You really feel like you can fall off when you approach the cliff. One thing I found interesting is how your body is tricked into reacting to things when you accidentally bump into them. It feels suddenly like there is something real there. This demo also shows how high quality shading and lighting can mitigate some of the shortcomings of the headset such as the resolution and screen-door effect. Another demo I tried, 4thFlrStudio by Brendan Coyle demonstrates this clearly, with its remarkable lighting and detail.

After a bit of messing about I got Steam and Half-Life 2 working on the Rift. Within minutes it gave me motion sickness! It clearly showed how VR requires both a rock-solid framerate and well-calibrated motion settings in order to be effective. Anything less than 75fps will likely send the user running to the toilet!

The presence of other characters in the game is startling. This was the first time I had encountered another being in VR, albeit a simulated one. At the beginning of HL2 you go through a police checkpoint and one of them thumps you in the chest. I actually felt it! The enemies were extremely intimidating. I dread to think what it is like on the later levels when you encounter the Hunters and Striders! I really felt I was inside the game, it’s kind of like a fuzzy lucid dream. I will have to experiment with the settings to see how far I can minimise the motion sickness.

The next thing for me to do is get the SDK going and see what I can create with it. I might also need a more powerful computer. Some of the demos I tried only ran at 16fps. My 2012 Macbook Pro is a great for general use and for my iOS development, but its NVidia 650m has about 1/7th the power of a top desktop GPU.

As you might already know my forthcoming iOS game, Toxin involves shooting rapidly dividing toxic cells which eventually fill the playfield or burst to spawn flying enemies.

Now, not long ago I read an article about generating heat maps to display web statistics, and I thought it would be cool to use this technique to colourise the ever changing clumps of cells in Toxin. After finding out how they work, I knocked up a prototype in Processing before adding the effect to the game.

Generating heatmaps is pretty simple. Take a 2d grid of values, perform a Gaussian blur on it, then use each resulting value as an index into a gradient of colours.

For Toxin, the first step was to create a simple array of numbers representing the current state of the cell grid. Toxin cells can be different sizes. To simplify things, I store each big cell as several small cells.

Each cell is given an initial value of 255. When we draw our cells, we use a range from black to white, where values of 0 are black, 255 is white and values in between are shades of gray.

The cells are represented as a grid of numbers. Each cell is given the initial value of 255

The next step is to apply a simple box blur to this grid of numbers. Normally heat maps use Gaussian blur, but since Toxin’s grid is such a low resolution we can get away with using a simpler algorithm which is much faster to calculate. To learn more about blurring techniques, I recommend Alex Evan’s article on Gamasutra.

The cell grid after being blurred. Notice how the values have changed

Now look what happens when we draw our original cell grid using the values from the blur operation. Cells in the middle of large clumps have higher values than solitary cells.

Drawing the original grid with the blurred values

The final step is to colourise these grayscale values so they look cool. We do this by mapping these values onto a gradient I made in Photoshop. Values of 0 map onto the left of this gradient, 255 to the right, and the others to the colours in between.

When we draw the grid with the colours from the gradient we get our heatmap!

For some reason the front end design is something that I absolutely must get right before I can envision the whole game. At the start of a project, I will spend hours trying to make a title screen and find the right font. If I get that right, everything else will follow. The design elements that I choose for the title screen seem to become the foundation for the graphical style of the whole game.

Needless to say, Toxin’s title screen, like everything else in this supposedly simple abstract shooter was a struggle to get right. It was really just lucky experimentation that lead me to the nice animated effect I use for Toxin’s title screen. This effect warrants a detailed blog post of its own, and I’ll write one soon. Today I want to write about how I made the buttons for the main menu using a 19 year old graphics program.

Toxin’s main menu. What you can’t see from this screenshot is how the glowing colours shift and change using multiple blended layers.

The icons on each button are simple pixel art images which were vectorised, loaded into 3D Studio Max, then extruded. My first icons were made by extruding text using the Topaz font, which was the default font on AmigaOS back in the day. I found it impossible to signify everything using text glyphs alone, so I decided to do some pixel art.

I found Photoshop too frustrating to use for pixel art. It’s just not geared for that kind of work. You can’t delete quickly, and when you’re zoomed in the cursor does not snap to the pixel grid. So I decided to go full oldskool – I started up my Amiga Emulator, WinUAE, and loaded Deluxe Paint 4.

In a couple of minutes I had knocked up some simple 16*16, 1 colour icons ready to be imported into 3DS Max.

DPaint 4 running on the WinUAE Amiga emulator. Drawing sprites inside a 16*16 checkerboard is an old technique from the Amiga days

For those who don’t know, Deluxe Paint (invariably referred to as DPaint) was the de-facto standard graphics program on the Amiga computer, just like Photoshop is for today’s machines. It was also the industry standard for video game art up until the start of the 3D era. (At least it was in the west. I have no idea what was used by the Japanese games industry to create such pixel art masterpieces as Seiken Denetsu III and Final Fantasy 6. If you know, send me a message).

It was developed by EA, believe it or not, back when they released interesting creative software like the Deluxe Music Construction Set, and published games like The Bards Tale. The first version was released in 1985, and was in development until 1994, which was probably the beginning of the end of the Amiga era.

I still have my copy of DPaint I. I bought it for £5 in 1991 when I was a poor schoolkid saving my dinner money to buy games. By that time it was already six years old and two versions behind, but I spent hours on it drawing cyberpunk cityscapes and trying to copy Street Fighter characters from magazines.

Creating good looking graphics for an abstract game like Toxin is harder than it first appears. Not only are your images stripped bare of most signifying information, requiring a real eye for the principles of graphic design, but creating the artwork itself – giving each line the right balance of sharpness and glow, demands careful image processing and a great deal of experimentation.

When I started making Toxin, the graphics were a real struggle. It took me an enormous amount of work to develop the look of the game. To give you an idea, I have a folder on my laptop with all of Toxin’s artwork in, and it weighs in at 967meg! All for a little mobile game!

Eventually all my experimentation began to pay off, and I developed some powerful techniques for creating animated multicoloured abstract sprites. In this post I want to describe one of my favourite methods.

I start by creating our sprite out of spline shapes in 3D Studio Max. Unlike many 3D programs, Max can render splines at various thicknesses, animate them, and can even apply texture maps and materials to them.

I then use gradient maps to mask out areas of the shapes that I want to be different colours. These gradient maps are then animated, so the different coloured areas can actually move around the shape.

Then, I do two sets of renders. First I render the shape with the masks on, so there’ll be a hole wherever the coloured area will be. Then I do another render with the mask inverted, so that only the coloured area is visible.

This diagram shows each step in the creation of one frame of animation for a Toxin sprite. The total animation is 64 frames long. This intricate process is largely automated using Photoshop scripts

Then its time to start up Photoshop. After deciding what colours I am going to use, I run each set of sprites through a Photoshop script that colourises them and adds a glow. To create the glow, I duplicate the image into another layer, Gaussian blur it, then additive-blend it on top. Sometimes I use several glow layers with different Gaussian blur parameters. I also find that using image->adjustments->Hue/Saturation to colourise gives me the best results.

The final step is to compose the two sets of colourised renders into a single sprite. I do this with another Photoshop script. The end result is a lot more interesting than the basic Geometry Wars style sprites everyone else is using, I think.

In game development it is common to have hundreds if not thousands of texture maps and animation frames in a single project. Keeping track of all these images is taxing for both the developer and the computer, so what we do is arrange the images into texture atlases.

A texture atlas is an image that contains many other images. Usually, all the animation frames for a single sprite, or all the textures for a single object are arranged into a single image. The game code then looks at a smaller portion of this image when it needs to draw a particular sprite or texture.

AtlasMaker is a Photoshop script that lets you create these atlases inside Photoshop. It takes a directory of individual textures and arranges them on a single image. It can be used both for atlases of different sized images, and for tile grids which are commonly used in 2d games. Earlier versions of this script were released in 2009. I know I promised an update long ago, but I have been working on my iPhone game, Toxin. I have been using this script throughout Toxin’s development, but never got round to tidying it up for public release.

AtlasMaker main window. A tile grid for a 2d game is being created

Features

• Cross platform – tested in Photoshop CS3,CS4,CS5 on Windows and Mac OSX.
• Open Source
• Create texture atlases or tile grids for 2d games.
• Several image sorting algorithms. Find the most efficient one for your textures.
• Add a margin to each image.
• Custom data file export.
• Extendable – It’s easy to add your own rectangle packing algorithms and sorting methods.

Generated with AtlasMaker: Left, a texture atlas with variable sized images. Right, a tile grid. Images used are randomly generated test textures.

Installing AtlasMaker

There are two ways to running AtlasMaker in Photoshop.
Unzip atlasmaker-v0.7.zip to your photoshop scripts folder. On windows this is usually: C:\Program Files\Adobe\yourphotoshopversion \Presets\Scripts
On a Mac this folder is at: Applications/yourphotoshopversion/Presets/Scripts

When you next start Photoshop, AtlasMaker should appear in the File->Scripts menu.

Alternatively, you can run the script without installing by unzipping the atlasmaker folder somewhere, then selecting Scripts->Browse from the file menu and then selecting AtlasMaker.jsx.

Quickstart guide

The first thing to do is select a directory of images by clicking “browse” at the top of the window. Once you have done this, AtlasMaker will scan through the images and collect size information about them.

Next, you want to select your packing method. If your images are texture maps of different sizes then you want to select the “Atlas Maker”. If you are making a traditional 2d game where the sprites are all the same size, then you want to select the “Tile Grid”

You’ll notice that some text will appear underneath “Number of Files”. This is a notification from the Tile Grid Packer telling you how many rows and columns your images will take up given the default document size. Different packing methods provide different notification messages according to their nature.

Then you can optionally select a sorting method. Sorting the images in different ways can improve the efficiency of the texture atlas. Some packing methods do not allow sorting, and will disable this option if they are selected.

You can also add a margin here if you want a gap between your textures. Margins are added on to the width and height of each image just like CSS margins.
Next, click “Document Settings” and you will be able to set the size of the texture atlas you are going to create. You can also set the document name here, and choose if you want to flatten all the layers into one, once the atlas is complete.

Now click “Create Atlas” and you’re done.

Custom data file export

One of AtlasMaker’s most powerful features is the custom data file export. You can export a text file for each texture atlas containing information about each image on the atlas. This might be XML or JSON to be loaded by a game engine, or you could use it to directly generate source code to be pasted into your application.

AtlasMaker custom data file panel. Here an XML fragment will be generated for each texture in the atlas. filenames and texture positions are substituted into the text using tags

Tags are used to substitute information about each image into the text:

• #i – Image index (0.. number of images in directory)
• #filename – Filename of source image
• #width – image width
• #height – image height
• #x – X position of top left corner on atlas
• #y – Y position of top left corner on atlas
• #p – page number

If you click the “Reorder export file” button, you can rearrange the order in which textures are listed in the export file. Just select filenames from the list and click “up” or “down” to reorder them. You can select multiple filenames by shift-clicking.

The zip file contains a readme.txt which describes every option in detail.

AtlasMaker is designed to be extensible. It is easy to add new packing methods and sorting algorithms. I have written the code to be more or less self explanatory in this regard, but if you want me to write something about it, please let me know.  And let me know if you find any bugs!

Download the latest version of AtlasMaker from Github

Download from Github

Processing provides a simplified front end to Java, ideal for quick experiments

The green cells in Toxin were all generated using a Processing program. Each cell is a circle made up of a number of connected spline segments. I animate the cell by displacing the spline control points using Perlin noise. The cells are then resized and colourized in Photoshop.

Left, a cell being rendered in Processing. Right, the finished result

As you can see, the finished product does not look exactly like the original Processing output. I had to do a lot of tweaking, re-rendering and reprocessing until I got them looking right. It was very time consuming. There were many moments where I wondered if I was going to get useable results at all.

Many of Toxin’s graphical effects have their origins in experiments I did with Photoshop blend modes. I discovered that certain combinations of layers and modes resulted in nice abstract animations when the layers were dragged around. I wrote photoshop scripts to perform these movements and spit out animation frames. Unfortunately these scripts revealed many bugs and inaccuracies in Photoshop and its scripting system. You only have to look through the source of my AtlasMaker Photoshop script to see how many strange workarounds are required to accomplish straightforward tasks.

To solve these problems I implemented Photoshop’s blending modes in Processing, and use this to create my animations. Below are a few frames from Toxin’s shot animation that demonstrate the technique. The way the colours shift throughout the animation is down to the movement of multiple layers blended into a simple white shot sprite. You can get a better idea of how this looks in the game by looking at the screenshots I posted a couple of weeks ago.

Many of Toxin’s sprites use animated image processing effects

I also used Processing as a test bed to work out algorithms before implementing them in the game. For example, Toxin has a weapon which freezes the edges of groups of cells. It didn’t take long to prototype this in Processing.

Testing a powerup that uses a flood fill algorithm to find edge cells

Stay tuned for more posts on Toxin, and a new version of my texture atlas Photoshop script which is in final testing…

Also, Toxin is the first game I have worked on since the Smartphone stuff I did around 2002-2003. I was desperate to get back into game development and I mistakenly thought I could begin again where I left off. It was like running into a brick wall. The first artwork I did was so bad I deleted everything, took the bus into Stratford on Avon and spent the rest of the day sitting by the side of the river in despair!

It took me some time to reach my old level again and then improve on it. As much as I have enjoyed developing Toxin and the creative discoveries I have made, there has been a great deal of struggle and fruitless experimentation involved.

The interior designer and UK TV legend Laurence Llewelyn-Bowen was an unlikely influence. On his TV home makeover shows he’d get his clients to collect images they liked and pin them to a board for inspiration. I had a Toxin folder on my laptop which I filled with images from ffffound and other design inspiration sites. When I got stuck I’d fire up IrfanView and spend half an hour blanky staring at images until some ideas popped up. I highly recommend this practice. A few of the images from my Toxin folder are in the image below.

That watch is a Dievas Aqualuna Blue Professional. No I can’t afford one yet. Buy Toxin!

The first thing I did was the main ship, which is the only thing I got right first time; it remained unchanged throughout development whereas everything else got remade at least three times! I think it works really well and has a nicely balanced, iconic look. I designed it in a vector drawing package which I later abandoned in favour of using rendered, animated splines created in a 3d program.

The background was next. It was essential to get the background style nailed as it would define the look of the whole game. I tried a few different styles as you can see in the rough drafts below.

One of my biggest goals here was to create something that looked good within the phone. Too often, mobile games just present windows on to the game world without any consideration for the device that is running the game. To me this is like designing a watch face without taking the case into account.

Next time I’ll talk about how Toxin uses procedural graphics and describe how I used Photoshop scripting and Processing, a Java based generative arts platform to create much of the game’s artwork.

Toxin is an abstract puzzle/shooter for the iPhone and iPod Touch, running on iOS3.0 and above. The basic gameplay is simple. You control a ship that rotates around the edge of an elliptical playfield, shooting inwards. If anything touches the edge you lose energy. Lose too much energy and you die.

Inside the ellipse are 99 levels of swarming biological foes, multiplying toxic cells, physics puzzles and colourful particle effects.

Toxin was inspired by a number of different games including Geometry Wars, Tempest, Space Invaders and Ebonstar, but the biggest influence was Microbes, released by Vision Software in 1991. I really enjoyed this back in the day, and like many innovative Amiga titles it seems to have disappeared from the consensual history of video games. You can play it with UAE, but I dont think the emulation is perfect.

I also tried to add some simple puzzle concepts to give the game more depth but without affecting the pace too much. The design process involved me sitting down with a coffee every night after work, staring at my sketchpad and thinking, “now what can I do inside an ellipse?”

Once this game is done, I am thinking about writing a Toxin “Remix” called Antigen. It will have the same ship-in-an-ellipse structure but the core game will be more puzzle based and slower paced as well as having different artwork. I think the idea of a game remix could work well at low pricepoints, a little like the way remix CDs used to accompany singles.

Soon I’ll write a little on how the artwork was done, and how it went through a long, painful process of development. I am also looking into producing a gameplay video, so stay tuned…

This is the titlescreen of Toxin, my first iPhone game. It was only supposed to be a warm-up; a short project to help me get back into the game after a few years developing web applications, but it has taken much longer than I anticipated.

I’m doing the whole thing myself, code, graphics and sound, largely because I don’t know anyone who could help me out, but also because I was a child in the 80’s, and despite all the developments in gaming that have happened since then, I still have the image of the lone game developer somewhere in the foundations of my mind. I grew up idolizing 8-bit legends like Tony Crowther and Andrew Braybrook, and learned to code in a world where games had authors associated with them. Things have changed, but I still carry all that with me; I just wouldn’t feel like a proper game developer if I didn’t do at least one game on my own.

Oh, and the title screen is animated. It uses a pretty cool effect that I will probably write about soon.