Game Development Stack Exchange is a question and answer site for professional and independent game developers. It only takes a minute to sign up.
Sign up to join this community
Say I have a framework with characters that can run and walk and of course, stand still.
Am I right in thinking that animation blending is the process of smoothly transitioning between these by generating dynamic animations at runtime?
If so, what is the basic theory of how this works, and how would I implement it?
Character animation blending is more general and can be used for:
The basic idea is that for two identical skeletons with different animation clips, for each joint rotation, you compute in-between values. Animation clips are represented with rotations. So linear interpolation is not what you should use if you want accurate results (even if you are using Euler angles). You can use slerp for example but more complex solutions exist.
This is a complex problem with complex solutions. Basically, if you have two animations clips A and B:
If you are interested, I suggest you start by reading this article so that you can better understand what the difficulties are: http://graphics.cs.wisc.edu/Papers/2003/KG03/regCurves.pdf
It is not the most up to date one, but it is easy to read.
I hope this helps. Best regards.
First off, I assume you are talking about skeletal animation (especially 3D graphics) since animation blending pretty much only applies to skeletal animation. If you are talking about 2D sprite-based animations (you didn't specify in your question) then you can pretty much forget about animation blending.
With that out of the way...
Am I right in thinking that animation blending is the process of smoothly transitioning between these by generating dynamic animations at runtime?
That is pretty much what the term means, yes. That phrase "generating dynamic animations" is a tiny bit misleading however; that brings to mind procedural animation and that isn't exactly what you're doing.
"Animation blending" is when you transition between two animation clips by using a bit of both animations with the weighting of one clip going up while the weighting of the other clip goes down. This works best with skeletal animation, so that the rotation of each joint is set to a midpoint between that joint's animation in two different animation clips, with the weighting of each clip changing over time. That is, you'd calculate the rotation of each joint something like:
rotCombined = weightA*rotA + weightB*rotB
Now, while animation blending works best with skeletal animation, you can also do it with vertex animation (eg. MD2). This is more of a historical note because few modern game engines use vertex animation, but you can set the vertex positions in the same way, as a weighting between the vertex position in two different animations. This could result in some weird looking interpolations (eg. parts of the mesh collapsing or turning inside-out) if the two animations were too different, whereas skeletal animation ensures that the various parts of the mesh stay whole.
