2009-10-07, 10:58 PM
I think one thing to bear in mind,
The move object animations are really just simple transformations (translation/rotation) ...the animation of monsters/npcs is much more involved whatever techniques are utilized. There are a number of different ways models are animated in this fashion.
The two major ways are either just interpolating between beginning and ending states (basically unique geometries within themselves) with possible intermediate frames (again unique geometries) ...usually this interpolation is linear, but it need not be, and individual curves can even be applied to each vertex. This animation can be global or local to groups of vertices.
The other approach is to have a kind of skeleton, which can either be pivot points or vertex groups. Individual transformations (translation/rotation/scaling) are applied to these groups. This can be very simple with skeletal like figures, or more complex with "skinned" figures. Probably the most common modern approach for real-time applications is to assign weights to each vertex and compute the final position of the vertex based on the weighted sum of usually up to four such pivots (transformation matrices)
It's very possible SOM incorporates both basic approaches. I think modern weighted skinning might've been a little much for the time / hardware / general level of technology demonstrated by SOM. But I would not rule it out or anything else.
The move object animations are really just simple transformations (translation/rotation) ...the animation of monsters/npcs is much more involved whatever techniques are utilized. There are a number of different ways models are animated in this fashion.
The two major ways are either just interpolating between beginning and ending states (basically unique geometries within themselves) with possible intermediate frames (again unique geometries) ...usually this interpolation is linear, but it need not be, and individual curves can even be applied to each vertex. This animation can be global or local to groups of vertices.
The other approach is to have a kind of skeleton, which can either be pivot points or vertex groups. Individual transformations (translation/rotation/scaling) are applied to these groups. This can be very simple with skeletal like figures, or more complex with "skinned" figures. Probably the most common modern approach for real-time applications is to assign weights to each vertex and compute the final position of the vertex based on the weighted sum of usually up to four such pivots (transformation matrices)
It's very possible SOM incorporates both basic approaches. I think modern weighted skinning might've been a little much for the time / hardware / general level of technology demonstrated by SOM. But I would not rule it out or anything else.