zAnimation – a simple XNA animation library

Before XNA 3.1, there was a bunch of animation frameworks which could to tons of stuff like interpolation between animations and stuff like that. When the 3.1 came out all of these became obsolete – except kw animation which comes with a 3ds max exporter and is great if you’re using max, but I’m using Maya and I needed to write my own animation foundation. zAnimation is a XNA library for animating skinned models exported from Maya. It supports customizable pre-clip interpolation, and per-bone animation. No interpolation between frames or animations yet.

Pre-clip interpolation:

Before a clip is played, the model interpolates over a customizable number of frames to the starting position. It delays an animation a bit but removes jerky transitions between animations.

Per-bone animation:

When defining a clip within the animation, you also define what bones does the clip affect. Based on that you can play an animation that animates all the bones in a model, and easily override the animation on one or more bones to make them move in a different way.

Below I’m listing a couple of things that I encountered while writing zAnimation:

Maya gets all your animation keyframes and exports it in a single take called “Take 001”.

To deal with this I introduced another file (the take definition file) which contains where does a certain animation start and end, and also what bones does it affect.

XNA FBX Importer removes redundant bones.

If a bone has no weights defined on a skin it gets deleted. The same thing happens if the bone doesn’t move (change) in the keyframes. So I introduced 2 special parts of the animation, the “Bind” part and the “Init” part. The bind part acts like a starting transform, and the init part is just there so you don’t forget that you have to animate a bone if you want to have it.

So based on that there are a few rules and limitations to zAnimation:

– smooth binding only

– each bone in each skin has to have some weights, so no root bones with no weights (or it won’t be imported as a bone)

– each bone has to have at least 2 keyframes that are not the same (or it won’t show up in the keyframes at all)

– strictly defined animation timeline:

-at frames 0 and 1 keyframe every bone at its bind pose position. This is the position your animation starts with.

-at frame 3 you should keyframe every bone it any position other than its bind pose. I select all my bones and rotate them in the x-axis about 45 deg.

-at frame 4 you should keyframe your bind pose again.

– after frame 4 animate your model as you wish

– when importing a model create a takes.def file in the same directory as your exported fbx is in the XNA content folders, set its build action to none from Visual Studio

– takes.def file has a specific structure that is line delimited, so the first line should look like this:

Bind, 0, 2, AffectedBone0, AffectedBone1, AffectedBone2

and each line after that should look like this:

AnimationName, StartFrame, EndFrame, AffectedBone0, AffectedBone1, AffectedBone2

here’s an example of a takes file

Bind, 0, 2, root, door
Open, 4, 34, door
Opened, 34, 36, door
Close, 36, 66, door

That’s more or less it. I’ll release it on codeplex in a couple of days. Check it out if you’re a XNA developer working with Maya.

Abandoned: Terrain

I’m building a little game called Abandoned, and I’ve just taken the first step and built the terrain. I used Shawn’s excellent article about detail textures found here. It’s just a simple terrain, but the colors and the fog settings have a huge impact of the atmosphere in the game. Even though it’s flat, I’ve built the vertices and indices as a little grid, maybe I decide later to have a heightmap or something. I’ve encountered one problem so far, and that’s poor texture quality. Then I remembered something about mipmaps and turned on their generation on the content pipeline options, and the problem disappeared. The result:

I wish to share my terrain generation function that I’ve been using for a while now (I probably stole it from somewhere, I just can’t remember where). It’s good for heightmaps and flat terrain alike. The size is an int because its the number of rows and columns, size of CellSize in the terrain grid.

void Build(int Size, float CellSize)
{
Debug.WriteLine(“Building ground vertices and indices. Size: ” + Size + “, CellSize: ” + CellSize + “.”);
vertices = new VertexPositionTexture[(Size + 1) * (Size + 1)];
indices = new int[Size * Size * 6];
for (int i = 0; i < Size + 1; i++) { for (int j = 0; j < Size + 1; j++) { VertexPositionTexture vert = new VertexPositionTexture(); vert.Position = new Vector3((i - Size / 2.0f) * CellSize, 0, (j - Size / 2.0f) * CellSize); vert.TextureCoordinate = new Vector2((float)i / Size, (float)j / Size); vertices[i * (Size + 1) + j] = vert; } } for (int i = 0; i < Size; i++) { for (int j = 0; j < Size; j++) { indices[6 * (i * Size + j)] = (i * (Size + 1) + j); indices[6 * (i * Size + j) + 1] = (i * (Size + 1) + j + 1); indices[6 * (i * Size + j) + 2] = ((i + 1) * (Size + 1) + j + 1); indices[6 * (i * Size + j) + 3] = (i * (Size + 1) + j); indices[6 * (i * Size + j) + 4] = ((i + 1) * (Size + 1) + j + 1); indices[6 * (i * Size + j) + 5] = ((i + 1) * (Size + 1) + j); } } vb = new VertexBuffer(GraphicsDevice, (Size + 1) * (Size + 1) * VertexPositionTexture.SizeInBytes, BufferUsage.None); ib = new IndexBuffer(GraphicsDevice, 6 * Size * Size * sizeof(int), BufferUsage.None, IndexElementSize.ThirtyTwoBits); vb.SetData(vertices); ib.SetData(indices); Debug.WriteLine("Finished building ground vertices and indices.");} [/sourcecode]

zCamera on codeplex

It’s all here: http://zcamera.codeplex.com.

Please post your feedback either here, on codeplex, or on my email.
Looking for XBOX developers for testing and rewriting the code for XBOX.