Yeah speedtree uses the APEX Vegetation module and does quite a good and efficient job of animating foliage by applying dynamic real-time phsyics modelling to them. While it does this sort of modelling (either for pre-recording or real time animations) quite well. Havok Animation lets you do something that allows you to ensure a realistic range of quickly generated and quickly executed animations as well as a way to intelligently propogate that animation across a population of models. With humanoid models Havok animation would let create crowd animations like "the wave" by assigning position based relationship between the models' wireframes. By having one model stand up and start the wave, the system would very efficiently propogate that initial movement across a population of very diverse looking models as long as they had sufficiently analogous wireframe components (do not actually need to have even remotely similar initial posing of that wireframe though.)
Through a comination of mirroring and blending Havok can translate animation across a set of models in directionally biased (like say the direction of initial motion/wind/the wave) and realistically fading way. Gusts of wind could be shown to blow across across a dense treetop canopy at greatly reduced computational costs because the work done in generating the initiating animation(s) will constitute much of the computation for all subsequent triggered animations. The same animation package that will be used to animate creatures from skeletal wireframes also includes computationally efficient ways to set of releastic foliage animation to emulate their response to steady and gusty winds as well as to animate folliage in response to the beating wings of a dragon.
By linking tree wireframes to components of a dragon's wireframes (most notably it's wings) you can have foliage that is disturbed and set into motion that will be appropriate to the frequency of its wings beating, their proximity and the direction from which the movement appears to propogate. This is all somethign that can be done by blending and mirroring a small number of animations rather than performing even the most rudimentary simulation of such a complex physical system.
It also would be more consistent and - even in terms of work to pre-render such scenes, would demand far less input and computational time to produce in large populations of varrying model density. What this requires, and what I suspect they developed, is a forect and foliage generation system that establishes the linkages as the forest is placed or generated. That would give them a tree generation system that utilizes their existing wireframe animation package to create forests effectively know to act like they are being blown in the wind, illustrate rolling and propogating gusts of wind, or sway and bend violently as a dragon beats its wings amongst them in proportion to the speed of its wings beating and their proximity to them.
Looking at the examples on the Havok site demosntrating animation belnding in real time between apparently dissimilar humanoids and looking over the demo of the software - this would be a very doable and shockingly computationally light way to animate massive clusters of folliage:
1) Utilizing the same physics package used to handle character animation
2) That can animate in in real time response to the animation of select model types (ie dragons blowing trees and people parting grass but no rats making willows snap) and to do so in proportion to the magnitude, frequency, and distance of those models.
3) Require less total time to develop animated folliage models by extrapolating a wide array of realistic movements from supperposition of easily modelleable and controllable normal oscillation modes (belnds simple oscillations of semi reigid skeletal frame into wide range of smooth realistic tree behavior)
All of this almost demands you develop a new method with which to generate not just tree models but folliage clusters. This is ONLY efficient if linkages are intelligently assigned automatically to some large degree. If you have to define those relationships manually all and always or have object representations that lack the bone-structure-like wire-frames then you would spend months of time tracking down unliked or improperly linked folliage - which would present as improperly moving or non moving loners or groups amongst the rest. These would stand out like secret service agents trying to blend in on a crowded and rowdy dance floor.
What I'm excited about all this though is that even if Bethesda does not use the Havok Animation features in this way - using the Havok Animation package opens up so many possibilities for realistic but computationally-cheap crowd animations and all sorts of things that would have been impractical in their older titles. If they are handling LOD as well as they seem to be bragging about - you should be able to see and do some things with large numbers of NPCs that would have flat out crashed their previous games in seconds.
No it will not make 20,000 watermelons dropping onto the head of the annoying fan npc resolve any more smothly but it does mean you could potentially animate a marching army realistically and smoothly - with behavrior such as the first and last marcher in a long column being slightly ou of step with each other and that phasing propagating and blending realistically from first to last soldier marching.
Through a comination of mirroring and blending Havok can translate animation across a set of models in directionally biased (like say the direction of initial motion/wind/the wave) and realistically fading way. Gusts of wind could be shown to blow across across a dense treetop canopy at greatly reduced computational costs because the work done in generating the initiating animation(s) will constitute much of the computation for all subsequent triggered animations. The same animation package that will be used to animate creatures from skeletal wireframes also includes computationally efficient ways to set of releastic foliage animation to emulate their response to steady and gusty winds as well as to animate folliage in response to the beating wings of a dragon.
By linking tree wireframes to components of a dragon's wireframes (most notably it's wings) you can have foliage that is disturbed and set into motion that will be appropriate to the frequency of its wings beating, their proximity and the direction from which the movement appears to propogate. This is all somethign that can be done by blending and mirroring a small number of animations rather than performing even the most rudimentary simulation of such a complex physical system.
It also would be more consistent and - even in terms of work to pre-render such scenes, would demand far less input and computational time to produce in large populations of varrying model density. What this requires, and what I suspect they developed, is a forect and foliage generation system that establishes the linkages as the forest is placed or generated. That would give them a tree generation system that utilizes their existing wireframe animation package to create forests effectively know to act like they are being blown in the wind, illustrate rolling and propogating gusts of wind, or sway and bend violently as a dragon beats its wings amongst them in proportion to the speed of its wings beating and their proximity to them.
Looking at the examples on the Havok site demosntrating animation belnding in real time between apparently dissimilar humanoids and looking over the demo of the software - this would be a very doable and shockingly computationally light way to animate massive clusters of folliage:
1) Utilizing the same physics package used to handle character animation
2) That can animate in in real time response to the animation of select model types (ie dragons blowing trees and people parting grass but no rats making willows snap) and to do so in proportion to the magnitude, frequency, and distance of those models.
3) Require less total time to develop animated folliage models by extrapolating a wide array of realistic movements from supperposition of easily modelleable and controllable normal oscillation modes (belnds simple oscillations of semi reigid skeletal frame into wide range of smooth realistic tree behavior)
All of this almost demands you develop a new method with which to generate not just tree models but folliage clusters. This is ONLY efficient if linkages are intelligently assigned automatically to some large degree. If you have to define those relationships manually all and always or have object representations that lack the bone-structure-like wire-frames then you would spend months of time tracking down unliked or improperly linked folliage - which would present as improperly moving or non moving loners or groups amongst the rest. These would stand out like secret service agents trying to blend in on a crowded and rowdy dance floor.
What I'm excited about all this though is that even if Bethesda does not use the Havok Animation features in this way - using the Havok Animation package opens up so many possibilities for realistic but computationally-cheap crowd animations and all sorts of things that would have been impractical in their older titles. If they are handling LOD as well as they seem to be bragging about - you should be able to see and do some things with large numbers of NPCs that would have flat out crashed their previous games in seconds.
No it will not make 20,000 watermelons dropping onto the head of the annoying fan npc resolve any more smothly but it does mean you could potentially animate a marching army realistically and smoothly - with behavrior such as the first and last marcher in a long column being slightly ou of step with each other and that phasing propagating and blending realistically from first to last soldier marching.
- Joined
- Nov 20, 2006
- Messages
- 1,710