We tend to split our purchases into categories, be that entertainment, identity, self-improvement, and so on. However, it turns that, most young people have a far larger budget for their identity purchases than entertainment. So if you can get your players to care how their avatar looks, in the same way they care how they look in the real world, they'll spend a lot more money buying avatar clothes than they would spend just buying access to new content, or extra games. For this reason, the style of your avatars is one of the most important decisions you will ever make about your world.

In this post we look at the two basic types of customisable avatar engines built into the Dubit Virtual World Platform, the Layering Engine, and the Cutout Engine.

Before we dig into the strengths and weaknesses of the two engines keep in mind that, although the example avatars are human, either engine could be set up to uses animals, or even vehicles, as avatars!

Layering Engine
Items of clothing are separate images layered on top of the avatar skin. The layers are composited into a single avatar image by the server. Each item of clothing is drawn and animated in all directions, usually North, South, West, South West, and North West. The different frames will be stitched into animation grids.

The layering engine produces the most detailed and graphically sophisticated avatars. It is often used for avatars that will be modelled in 3D, or require complex animations.

Although capable of beautiful avatars with complex animations, avatars produced for the layering engine tend to result in larger files that take longer to download. Consequentially, the layering engine trades graphical sophistication against file size, download time, and limits the number of animations.

We recommend that avatars produces for the layering engine should have no more than 4 animations, typically standing, walking, waving, and dancing.

Cutout Engine
The avatar is made up of separate body parts with the items of clothing matched to the body part. For example, a pair of jeans would be cut into 5 separate images, the lower left leg, lower right leg, upper left leg, upper right leg, and the waist. The server composites all pieces into one file.

The concept is illustrated by the Brit Chicks avatars shown below. The avatar is split into 18 different body parts, where each body part can be independently animated. The items of clothing are attached to the body parts and automatically follow the same animation path.

An avatar produced for the layering engine usually has eight image layers - skin through to hair and hats - and four animations - standing, walking, waving, and dancing. Because each frame of the animation is a separate image, and must be drawn in eight directions - North, Northwest, West, Southwest and so on - there are often over 500 individual images required to make up one avatar. Every new animation adds to the total file size.

Whereas, the cutout engine will usually have only four directions and require less than 50 images. Because the clothes are cut to match the body parts, adding a new animation is just a case of adding the body parts, the clothes follow.

It's difficult to model sophisticated 3D avatars for the cut out engine, instead the avatars are usually illustrated by hand. But, they can support far more animations, are quicker to load, cheaper to produce, and can easily be integrated into mini games.

Feature Comparison

Production Cost

We like to think of the Dubit Virtual World Platform as an empty virtual world excitedly waiting for lots of content, games, puzzles and avatars. It's this content that's going to make your world exciting and unique, everything else is just features and infrastructure. Lets face it, players won't choose your world because its got a unique buddy list, they'll choose your world because its got a great story, fun avatars, puzzles, adventures, and games!

So that you can focus on your content we've created tools for designing rooms, quests, and adventures, plus we manage the infrastructure, in our Amazon EC2 cloud, so that you don't have to.

The platform architecture looks a little like this simplified diagram.

You still host your website, art work, music, and the empty virtual world. The multiuser server, rendering server, web services, and persistence databases run in our EC2 cloud:

• Nexus Multiuser Server: built on top of the Apache Mina project, each node in our multiuser server cluster handles many thousand concurrent players. The multiuser server routes messages between players in the world, in extreme circumstances delivering over 250,000 messages per second.
• Ruby on Rails Web Services: responsible for proxying access to the MySQL database. The web services are consumed by the Nexus Multiuser Server and the Flash Client, returning profile, shop, product, and wardrobe data. For heavy lifting, the Rails services are backed up by Java services running on Glassfish.
• Ruby Admin Website: avatar clothes, furniture, and products are uploaded and managed through the admin website.
• Java Render Server: reduces avatar load time by compressing and combining assets into single files. Cuts avatar load times 10 fold.
• Persistence Database: anything that happens within your world can be automatically saved to the persistence database. For example, completing a quest, picking up an item, levelling up, unlocking a door, would all be transparently saved to the persistence database.

Last month saw our 10 year anniversary; that's a lot of years creating virtual worlds! Since we started we've designed our fair share of rooms. If there's one thing we've learnt, it's not to ignore the "Three S's" of virtual world design:

• Software: modelling a room in 3D software (we use Maya) simply takes longer than drawing a room into Flash or Illustrator, often 2 to 3 times longer! Compositing a room out of textures in photoshop falls somewhere between the two.
• Style: buildings are quicker to model and quicker to draw than organic shapes. The Real Life city shown below was far quicker to create because the shapes are geometric, where as the Cartoon Network island took far longer.
• Size: simply, the larger the room the greater the production time.

In nearly all cases a room needs 1 to 2 days of concept development, and between 0.25 and 2 days of integration into the world. But it's the Three S's that really determine how much work a room will take!

We've listed a six example rooms that illustrate the "Three S's" at work.

Matthew Warneford

Follow me on Twitter