Posted: 11/17/98    Previous | Headlines | Next

GameSpot News

Making a title like Half-Life look heads, shoulders and tails above the rest of the competition takes in-depth planning, lots of ingenuity, a die-hard understanding of computing, and a great deal of experience in the ways of how the human body works.

Ken Birdwell, senior software development engineer at Valve Software and its upcoming title Half-Life, has definitely amassed a hefty resume that definitely has placed him where he belongs.

Over the last 15 years, Birdwell has contributed to projects as hardcore as in-circuit emulators, 3D surface reconstruction, 3D prosthetics design tools, and satellite networking for Microsoft's Broadcast PC. On the gaming side, he wrote the first graphical front-ends for CompuServe's Sniper multiplayer gaming area.

While all of this sounds impressive to some, new gamers will definitely be playing with his latest work - the animation system for Half-Life. Birdwell designed and implemented the animation system and several other engine components that make Half-Life stand out amongst other Quake II-based games.

So what's his secret? GameSpot News caught up with Birdwell to find out.

GameSpot News: How is Valve using skeletal animation in the multiplayer portion of Half-Life?
Ken Birdwell: There are a bunch of different benefits. We can animate the different parts of the skeleton independently, so when you see your opponent in multiplayer, you can tell exactly where they are looking because that's where their head is pointed, which makes it a lot easier to sneak up on someone with the crowbar. You can tell exactly where they are running, because that's where their hips and legs are pointed. If someone is aiming up, their torso and head rotate back, and the weapon rotates to point in the direction they are aiming. You get a lot more feedback as to exactly what the other player is doing, and the great thing is, your brain already knows how to process what you are seeing without having to be told.

We wanted to address the problem of being able to see what weapon the other player was using. Just changing the weapon model wasn't really enough, we wanted to make it a bit more obvious so we have the player holding and shooting each type of weapon in a unique position: down low for the machine gun, over the shoulder for the rocket launcher, two-handed grip for the pistols, etc. We started out only going to do a few, but it looked so good that Doug Wood just went nuts and made around 100 different poses.

We can connect the weapon to the player's full skeleton. The Tau-Cannon is a good example. It has a big backpack component, a nozzle the player holds in their hands, and a hose connecting the two. Since it's connected to the player fully, when the player twists side to side and bends up and down, all the different parts follow their movement, just like you'd expect in the real world.

We also do a pretty good job of illuminating the player model. The easiest way to see this is to have someone go into a dark corner of a room and fire off the MP-5. Rather than illuminating the entire model equally, the hands and face will be brightly lit, and the feet almost unaffected (depending upon the weapon brilliance).

GSN: Is the player split into three parts - head, torso, and legs?
KB: We don't really break the model into independent sections. That would be simpler, but would result in flaws in how the pieces connect. Instead with Half-Life characters, it's always one model and we just modify the skeleton. We can blend up to 12 different animations per frame, eight in the upper body and four in the lower, though we actually blend the upper and lower body animations together through the spine as well, so it gets a little complicated. When we turn the body, we don't just turn the pelvis, we smoothly blend up the spine from the pelvis to the shoulders so there's no sharp disconnect, the model is always a solid continuous mesh.

GSN: Will players be able to customize their models?
KB: Initially we are releasing five different multiplayer models (Gordon, Gina, Helmeted Guy, Human Grunt, and Scientist). Players have sliders to control various aspects of their appearance. We're going to be releasing additional multiplayer models to the community on an ongoing basis. When you connect to a server that has additional models, they will automatically be propagated to your system, which also happens for levels, decals, sounds, or any other resources you need to play.

People can build multiplayer models that will use the existing multiplayer animations, which saves them a lot of time. They'll need 3DStudio, as well as our file exporter, but it should be fairly simple for anyone familiar with that program. We're going to be releasing all of the animations they'll need to build a completed player model, as well as all the other support files. All they really need to do is build a model with the same skeleton as our player model, and it just works.

Harry Teasley has already released a beta version of these tools to some animators, and as soon as they tell us it works for them, we'll be releasing them on the Internet.

GSN: Couldn't you cheat this way, make a model that is too small and couldn't be hit?
KB: This was a problem on some earlier games, but shouldn't affect Half-Life, at least until people come up with some other way to cheat. The multiplayer game always uses the default Gordon model to calculate damage. The user-created models are only visual and don't actually affect gameplay. Also, if another player is using a model you don't want to see, you don't need to install it, the game always defaults visually to the helmeted Gordon model if you don't have whatever model they are using.

GSN: Can the players customize the skin texture? How about shirt/pants colors?
KB: You need to build a new texture and regenerate the model to put the new skin on. We do allow for changing shirt/pants colors, but in a slightly new way. Rather than having the player pick from a few preset colors, we have them pick a hue they want. This allows for tens of thousands of combinations and lets them pick a very exact color scheme. We also look at more information on the model itself so that areas that are supposed to be bright and shiny looking are bright and shiny, and areas that are supposed to be flat and dull are flat and dull. It guarantees that whatever colors they pick will look right on the model they've picked.

GSN: What are your animation plans for Team-Fortress? Any major changes?
KB: Well, expect a lot more flexibility in the player models and a lot more variation of movement. We're trying to get the models to convey as much as possible about the what each player is doing, where they are looking, what weapon they are holding, their health, basically any state that is important to multiplayer. Actually, Half-Life multiplayer was the testing ground for some of the Team-Fortress features we were wanting. Robin Walker and Steve Theodore have taken the Half-Life work and pushed it to new heights, doing a lot of stuff I didn't think possible.

GSN: Thanks Ken.