I've mentioned before how important I think it might be to get "open source game cities" of the GTA sort available for use in many different kinds of projects...

Just recently I've been thinking more about a similar need for "open source avatar systems". It's easy to see that avatars have become nearly ubiquitous, particularly with Microsoft's, PS Home's, and Wii's avatars all used to varying degrees across multiple game projects. (Of the three, Microsoft's being the most accessible to independent development via the XNA platform.) Character creators and character customization can be a key differentiator today between big RPGs and the little guys. Even 2D games have customizable and reusable avatars beginning to be more common.

I think that a collaborative, highly customizable avatar system would be hugely useful if it were open source and available as a relatively easy API to embed into nearly any independent or open source game. Add in even basic sharing of avatars between games, and easy to use customization tools for both in-game player-oriented customization and out-of-game NPC development, and you could easily find a tool that many, many people would heartily use.

Of course, the blocking factor may not even be the technology so much as the art. Good customizable 3D avatars require complicated deformable models that I expect require a lot of smart tweaking from smart artists to get a good, wide, range of possibilities. That's before you start trying to build clothes and accessories that similarly fit the wide range of possibilities... 2D avatar systems are nearly as art intensive, as well I'm sure.

I started wondering if Second Life could at least provide some or all of an open source 3D avatar system. It seems like a decent place to start: there exists an open source viewer (with support for most OSes), and Second Life has a fairly wide variety of avatar models and avatar customizations that already exist, some of which already have CC licensing information attached. I am wondering what it would take to subsume "just" the avatar portions of the viewer into other game engines. Certainly it might be something worth investigating.