PDA

View Full Version : More Avatar Info - Rob Powers



kosmodave
04-23-2010, 08:32 AM
Apologies if this has appeared elsewhere but I found this interesting especially the Emanation part where they used character motion and a particle system to create the environment.

A Tutorial! ---- Rob? :D

Dave.

Link: http://www.awn.com/articles/technology/rob-powers-talks-vad-and-more-avatar/page/1,1

SplineGod
04-23-2010, 03:12 PM
You use the path extrude tool in modeler to use the characters motion path to rail extrude a branch. The motion path can be imported and convered to a spline. This spline can also be used with t he spline guid tool to deform a tube to the shape of the motion path. You can also use the spline to rail clone objects as well.
In layout the same motion path can be used along with the curve conform tool, DStorms trailer plugin etc to deform a tube( branch/root) along a path and then save transformed.
Particles can be generated on a ground object and plants, trees, etc fxlinked to them. You can also do the same with a partigon emitter, save transformed and use that along with point clone plus in modeler to distribute trees etc. :)

robpowers3d
04-23-2010, 05:51 PM
Larry,
This is not really what we did in the specific Avatar case. it really was more like a character performance driven near realtime process where Zoe's performance sculpted her environment. The first time that I know of this being done. Tree branch systems were actually created by the Actor motion and the geometry created was the template that determined what showed up in the final shots.

SplineGod
04-23-2010, 06:31 PM
Thats cool. :)
Ive used the method I described to generate geometry from mocap and motion data. :)

robpowers3d
04-23-2010, 09:10 PM
Yes Ive done that in modeler previously as well and although both techniques end up creating geometry the method I used on Avatar happened in the real time workspace and not in a separate modeling environment. I think that is the key difference.

NanoGator
04-23-2010, 11:03 PM
Yes Ive done that in modeler previously as well and although both techniques end up creating geometry the method I used on Avatar happened in the real time workspace and not in a separate modeling environment. I think that is the key difference.


As I recall, they captured that run based on what they could build on the stage that day. They didn't have an environment yet so they just captured the actors without the virtual environment and instead designed something interesting and tree'ish for them to run on. So, just like Rob's saying, the performance drove the environment, not the other way around.

Towards the end of the movie we had to use that process again. There's the montage where Jake was running amongst the roots with Neytiri, there's a line about his feet getting stronger. They did several takes of different running and climbing performances. Then the Lab blended those together into one big long motion. That changed where all the footfalls were, so the environment had to be rebuilt to accomodate that new motion. Rob's technique was used to lay the ground pieces down for that.

SplineGod
04-23-2010, 11:24 PM
So it wasnt something that was done in LW.
Maybe kosmodave was under the impression that this
was somehow being done in LW. What I was suggesting
were some ways to use the motion paths directly in LW to generate or effect geometry. :)

kosmodave
04-24-2010, 12:06 AM
Thanks for the additional info everyone,and yes Larry I did assume it was all in Lightwave. I have not had a chance to play with mocap stuff much but when I get a minute or two I will experiment with your technique Larry and see what I can achieve. Thanks again.

Dave.

NanoGator
04-24-2010, 12:10 AM
If you already have mocap data, the approach Rob described would do you wonders. You'd lay down footfalls then model around them.

kosmodave
04-24-2010, 12:28 AM
If you already have mocap data, the approach Rob described would do you wonders. You'd lay down footfalls then model around them.

Thanks again, Yes I will play with this idea as well. I have no specific need for these techniques right now but I would like to experiment anyway, You never know when the need may arise.

Dave.

SplineGod
04-24-2010, 02:58 AM
Dave,
You could also potentially parent a null to the bottom of a characters foot, bake the motion onto that and determine where on the spline the footfalls are making contact with a surface.
You could also copy/paste the geometry from the bottom of the feet into another layer and have that use the bones from the main character. Mdscan the feet geometry and use the editfx tools to derive a motion path from one of the points. Again use that to again determine where footfalls take place. :)

OnlineRender
04-24-2010, 04:57 AM
..


well done Rob 7 years to get a forum avatar :D