PDA

View Full Version : Should LW embrace parametric Animation Method? Pixar and Autodesk seem to like it.



robertoortiz
08-30-2015, 08:39 AM
Hi guys,
One of the hot topics that came from the SIGGRAPH 2015 conference (besides Virtual Reality, and Real Time Production Rendering with Game Engines) is how high end programs are embracing
parametric animation.
Hell in one of the lectures I attended Pixar mentioned that is the direction they are pushing their pipeline (starting with Finding Dory).
So what is parametric Animation?
From Wikipedia:
"Parametric animation is an animation technique used in computer software, such as in computer games, that blends two or more separate animations together to form a new animation. This new animation is constructed in real-time by the game engine, and is not stored in a separate file like a regular animation. The technique was first used in an early build of the Half-Life mod, Team Fortress 2, and it not only heavily reduces artist workload during game development, it provides for much smoother animation as well."
https://en.wikipedia.org/wiki/Parametric_animation

There are many reason for this , among them that companies with long libraries of character animation can reuses the data at will.
It makes calculation solvers for both cloth, environment easier and it lends itself to export content for real-time engine.
Since Lw is beign overhauled I think this is something that LW should look into as part of a new animation pipeline.

What do you guys think?

-R

jasonwestmas
08-30-2015, 09:18 AM
In this case the new technology being adopted is an obvious choice for improving efficiency. Re-purposing data has only one drawback and that is the overuse of re-using the motion. In which case the art becomes stale because of the lack of creative thought given to each asset as a unique creature or thing. That does not mean however that each and every asset cannot borrow or use something previously created as a solid foundation thus saving time and thus permitting change to the ingredients of a unique, textural experience within the animation.

In a manner of speaking allowing the computer to interpolate between two coordinates in space does allow for a smoother work experience but I view such estimations as a suggestion and not necessarily artistic truth.

But yeah in the light of all that is commercial, sometimes you just have to get it done! So the cookie cutter "best guess" approach can be your tweening partner who fills in the gaps for you quicker than you might.


BTW is lightwave being "overhauled"? I haven not seen anything to suggest this. Granted I'm only looking at this from a user's perspective.

stiff paper
08-30-2015, 09:44 AM
"Should LW embrace parametric Animation Method?"
No.

I genuinely can't bother myself to make a list, but it's quite easy to think of, say, 100 things that LW doesn't currently have that it really does need, and every single one of those things would be more important than "Parametric animation method."

Maybe when they've made a full set of good, usable UV tools, maybe when they've made good, usable way to easily make corrective morphs in Layout, maybe when they've added MIS, different GI methods on the first and second passes, an easily usable physically based material system, maybe when they've unified the deformer system to some extent, maybe when they've implemented some form of OoO or something analogous, maybe when they'd done the other ninety things on the list.

Maybe then.

erikals
08-30-2015, 09:49 AM
personally, yes, i'd love to see it


https://www.youtube.com/watch?v=_0zzXvjfomQ

jasonwestmas
08-30-2015, 09:49 AM
"Should LW embrace parametric Animation Method?"
No.

I genuinely can't bother myself to make a list, but it's quite easy to think of, say, 100 things that LW doesn't currently have that it really does need, and every single one of those things would be more important than "Parametric animation method."

Maybe when they've made a full set of good, usable UV tools, maybe when they've made good, usable way to easily make corrective morphs in Layout, maybe when they've added MIS, different GI methods on the first and second passes, an easily usable physically based material system, maybe when they've unified the deformer system to some extent, maybe when they've implemented some form of OoO or something analogous, maybe when they'd done the other ninety things on the list.

Maybe then.

well of course, that's a given lol. As well as layering animation and all that jazz.

MSherak
08-30-2015, 11:23 AM
Hi guys,
One of the hot topics that came from the SIGGRAPH 2015 conference (besides Virtual Reality, and Real Time Production Rendering with Game Engines) is how high end programs are embracing
parametric animation.
Hell in one of the lectures I attended Pixar mentioned that is the direction they are pushing their pipeline (starting with Finding Dory).
So what is parametric Animation?
From Wikipedia:
"Parametric animation is an animation technique used in computer software, such as in computer games, that blends two or more separate animations together to form a new animation. This new animation is constructed in real-time by the game engine, and is not stored in a separate file like a regular animation. The technique was first used in an early build of the Half-Life mod, Team Fortress 2, and it not only heavily reduces artist workload during game development, it provides for much smoother animation as well."
https://en.wikipedia.org/wiki/Parametric_animation

There are many reason for this , among them that companies with long libraries of character animation can reuses the data at will.
It makes calculation solvers for both cloth, environment easier and it lends itself to export content for real-time engine.
Since Lw is beign overhauled I think this is something that LW should look into as part of a new animation pipeline.

What do you guys think?

-R


Stated above is not a perfect solution and is nothing new. Lightwave has had the ability to mix animations for years with Motion Mixer plugin. You are not limited on what you can mix either. Setup what you like and mix to your hearts content. You can control all the way down to the independent channels for any item. Just takes setup time. Check the tools in the shed before looking to buy new ones..

Oh and yes you can use it as animation layers also.

erikals
08-30-2015, 01:39 PM
Motion Mixer...
i tend to forget it exists...

Ryan Roye - Delura
https://www.youtube.com/watch?v=I3zLvhWIiZs

Brian Williamson - BugZilla
https://www.youtube.com/watch?v=8bsuK2iduoc

William Vaughan - Proton
https://www.youtube.com/watch?v=bH7nNmA3_HI
https://www.youtube.com/watch?v=gDRchM7WEAY

robertoortiz
08-30-2015, 05:20 PM
Well a big part of the problem is the implementation JEBUS!
The implementation I saw from Pixar was simple sliders and expandable assets like AE.

jeric_synergy
08-30-2015, 06:50 PM
Holy Buddha, watching Ryan explain Motion Mixer was painful, in that it seems like a TERRIBLE tool. Far too twiddly.

jasonwestmas
08-30-2015, 07:38 PM
Let's be serious here, Motion mixer is terrible, which is why I wouldn't use it. Yes implementation is everything.

MSherak
08-31-2015, 01:02 AM
Well a big part of the problem is the implementation JEBUS!
The implementation I saw from Pixar was simple sliders and expandable assets like AE.


Holy Buddha, watching Ryan explain Motion Mixer was painful, in that it seems like a TERRIBLE tool. Far too twiddly.


Let's be serious here, Motion mixer is terrible, which is why I wouldn't use it. Yes implementation is everything.


HAHA. You guys crack me up.. Yes I am sure that 2015 Pixar motion/animation editing in their own custom tools looks great compared to a plugin that you have been able to use since version 6 of Lightwave when it comes to UI. They have a whole dedicated programming team just to make tools for their movies. Ed Catmull makes sure of that.. You know, the guy that invented Sub-D's.

For those that have not used Motion Mixer at all or watched a video, I think you are judging it before you try it. Animation blending is not easy yet Lightwave has had in for over a decade and the end results look great.. Yes, I am sure the user implementation (UI) could be improved. But of course one would have to use to know what areas need improving. Just looking at something else and saying I want that in what I use is silly. If this was the case then you would be asking for the edit abilities of Motion Builder for Lightwave way before the Pixar one. Motion Mixer will get the job done if you need that ability to blend animations into a new one despite the UI implementation. It will do exactly what you tell it to do. Everything exposed for the animator to use and control.

As for Pixar , they also do everything custom and this is why they are so flexible at making new cool tools. So showing the world some new animation tool they came up with is moot since they will be the only ones who get to use it. The only things Pixar gives out to the 3D community are file formats and now a render that has a ton of competition. Yes game engines have been doing this kind of animation blending for decades. If you want to see it's nothing new then look at http://www.radgametools.com/granny.html. It's a blending animation engine for games that has been around since the mid 90's and is used in so many games. FYI, started off with a Lightwave exporter back in the day.

Again don't say you won't use it cause it's terrible or the implementation is not like so and so. Use it and give constructive feedback, I bet you will be surprised at the results despite the implementation/twiddly it is.

jeric_synergy
08-31-2015, 08:25 AM
I'm not comparing it to anything but itself. I've >>attempted<< to use MM several times, and its ridiculous interface always defeated me. But if I persevered, and tried REAL hard, it'd crash.

It is NOT a good tool.

Ryan Roye
08-31-2015, 09:31 AM
The main reason I used motion mixer in the first place was the fact that I had no other way to blend motions and play animations relative to any given position in 3d space. Motion mixer was the only tool I was aware of that had that capability... until I discovered that IKBooster had them to some degree as well; just in a less interactive and way less obvious form. IKBooster is also not dependent on naming conventions like Motion Mixer is, therefor you won't get into those show-stopping paradoxical situations where you completely break one tool (motion mixer) if you use another (expressions/motion modifiers/plugins) and vice versa.

To be fair, I don't consider IKBooster the holy grail of motion management either... it just happens to be the only tool Lightwave has ever had that allows unrestricted relative motion placement (in any direction, not just X/Y or Z like motion mixer) and interactive adjustment to those relative motions, which is what I consider one of the most important things for a character animator to have access to. If you can't re-use motions in such a fashion, then inserting them into your existing animations to save time ends up becoming a very time consuming and inefficient process. Eventually, IKBooster's relative motion methods need to be replaced by something that is not only more user friendly (like motion mixer) and integrated better with the rest of Lightwave, but also be provided with SDK access to enable more advanced workflows to be built around that ability.

In regards to the original topic, parametric animation is indeed something useful to all 3d animators; it's a no-brainer.

jeric_synergy
08-31-2015, 09:46 AM
Dude, I'm just happy you were willing to take the hit of learning all the ins and outs of a terrible implementation.

robertoortiz
08-31-2015, 02:31 PM
The main reason I used motion mixer in the first place was the fact that I had no other way to blend motions and play animations relative to any given position in 3d space. Motion mixer was the only tool I was aware of that had that capability... until I discovered that IKBooster had them to some degree as well; just in a less interactive and way less obvious form. IKBooster is also not dependent on naming conventions like Motion Mixer is, therefor you won't get into those show-stopping paradoxical situations where you completely break one tool (motion mixer) if you use another (expressions/motion modifiers/plugins) and vice versa.

To be fair, I don't consider IKBooster the holy grail of motion management either... it just happens to be the only tool Lightwave has ever had that allows unrestricted relative motion placement (in any direction, not just X/Y or Z like motion mixer) and interactive adjustment to those relative motions, which is what I consider one of the most important things for a character animator to have access to. If you can't re-use motions in such a fashion, then inserting them into your existing animations to save time ends up becoming a very time consuming and inefficient process. Eventually, IKBooster's relative motion methods need to be replaced by something that is not only more user friendly (like motion mixer) and integrated better with the rest of Lightwave, but also be provided with SDK access to enable more advanced workflows to be built around that ability.

In regards to the original topic, parametric animation is indeed something useful to all 3d animators; it's a no-brainer.

Great post!

lightscape
08-31-2015, 11:32 PM
HAHA. You guys crack me up..
Again don't say you won't use it cause it's terrible or the implementation is not like so and so. Use it and give constructive feedback, I bet you will be surprised at the results despite the implementation/twiddly it is.

Can you show your results using it on a character animation in lightwave? Humor us.
Its seriously dated.

erikals
09-01-2015, 08:00 AM
as far as i see, Motion Mixer is best for background characters...

but hey (!) those are also required  http://forums.cgsociety.org/images/smilies/arteest.gif

MSherak
09-01-2015, 11:08 AM
Can you show your results using it on a character animation in lightwave? Humor us.
Its seriously dated.

I just used it on non-bipedal animations for work. 1650 transition in and out animations for 54 logos. I did the animation paths then stored them all as motion mixer files. Then just assigned the animations needed to the logos saved and sent to the farm. Took less than a day to get them all on the farm and rendering. Nice thing is if I need to re-time them or do another animation I can just place another animation in. Character animation is not the only thing that matters when it comes to motion blending.

I have used in the past week and is saved me a lot of time, there is your humor.

lightscape
09-02-2015, 05:25 AM
But most of the people who had problems with motionmixer are using it for more than spinning logos.
That's where the issue comes out.
I think the dev that originally created mm is long gone so the chance of updating it is nil.
So why not create a more robust mm tool or a new one?
There is just no reason to suppress such requests to improve lightwave.

jasonwestmas
09-02-2015, 08:00 AM
Everything I work with is rigged too soooo, motion mixer doesn't support such things. It's destructive to have to bake the bones and remove constraints every time I want to motion blend or layer something. Mixing and blending should be an experimental process not a linear one.

jeric_synergy
09-02-2015, 08:30 AM
Mixing and blending should be an experimental process not a linear one.
Especially since the whole concept behind Motion Mixer is "non-linear animation".

It was to be to animation what modern computer editing is to video-tape editing.

MSherak
09-02-2015, 01:21 PM
But most of the people who had problems with motionmixer are using it for more than spinning logos.
That's where the issue comes out.
I think the dev that originally created mm is long gone so the chance of updating it is nil.
So why not create a more robust mm tool or a new one?
There is just no reason to suppress such requests to improve lightwave.

They are not just spinning, but assembling.. Heck if it was easy animation per logo I would have just saved a motion file and loaded it back. I am not suppressing the improvement of MotionMixer, yes the plugin can be improved but yes we have the ability to blend animations currently too. Using MotionMixer correctly is 100x easier than exporting an .fbx and editing in MotionBuilder then bringing it back.


Everything I work with is rigged too soooo, motion mixer doesn't support such things. It's destructive to have to bake the bones and remove constraints every time I want to motion blend or layer something. Mixing and blending should be an experimental process not a linear one.

And there lies the problem. You need to have all the animation available to blend correctly. Which means you need to bake it down so there are no other influences. If not, getting MotionMixer to work within a rigged system means applying it only to the things that control the influences. You would have to this in any 3D commercial package. To experiment with non-linear animation blending, on a slider,panel or etc., you need a lot of animations to pull from, no way around that, baked or otherwise. Parametic animation is the result of two or more destructive baked animations, whole or localized, blended to create a new one on the fly. Not just blend this animation into another with all these plugins, ik, nodes, motion modifiers, etc. attached to it.


Especially since the whole concept behind Motion Mixer is "non-linear animation".

It was to be to animation what modern computer editing is to video-tape editing.

It is if the only data you are feeding it is animation channels which make it non-linear. Not plugin data that has not been evaluated yet till you scrub to that frame. Even simple things like reading the information of an item in a IK chain becomes and issue since that data is dynamic. Blending happens on static information, keyframes, tangible data. MotionMixer can blend this if you hand it that type of data. Even the ExternalXChannels in the Actor can be used so long data is written into that channel.



MotionMixer does not store anything more than hierarchy and motion data, keyframes for the items you select into a clip that can be re-targeted. IF you wanted this to work flawless inside of LW with all things active, plugins for example, then the underlying structure has to change. Which means the plugins would have to know they can have overrides coming in from other areas. Something that is not addressed at the lower level of the architecture of most animation systems, same thing happens in Maya.

I don't think Pixar's tools have to worry about plugins for animation, hence these issues do not need to be addressed compared to a commercial package with 3D party support abilities. It's to large of a subject to nail down to one solution at the moment. Even now to get an animation in a game engine you have to bake it to keyframes, no way around it. It's faster to work from fixed data. And at least at the 3D commercial level we can finally share animations between packages with .fbx. And in that format the basic structure can't have a influence on a animation channel no matter what 3D package you use to export.

Truly a better way to handle all this is in the future is move to point cloud feedback system. If we had Chronoscupt inside of LW it would be awesome. Tools like this are showing up all over the place in Maya. You can't move though the system without a bake this into a point cloud. It's much easier to blending point data then have to worry did that plugin bake it's animation correctly. Alembic is proof of that.

In conclusion, if you had 10 animations baked down to keyframes and loaded into MotionMixer you would find very easy to create a lot more parametric animations from that data. MotionMixer does what it claims. No more, no less. It's not just as easy hitting a button at the moment and everything works. It takes time and planning to setup. Question now is, where do we think the improvements need to be in MotionMixer within the current universal system of animation??

For me:
1) Non-scaling top buttons
2) Channel Editor panel floatable like Offset Editor
3) Actors list as a real list not a popdown so functions can be right click context (active, add, remove, etc.)
4) Motions list as a real list not a popdown so functions can be right click context (add, remove, load, etc.)
5) More Layers than just three
6) Properties panel on the right side of the window
7) Call Motions for Actors as Clips since they can contain more than one type channel from Layout.
8) Edit Actor Map have the ability to remove channels.
9) Different color for motion bar based on offset. (Blue ABS, Dark Blue REL, Dark Green REL XZ, Dark Red CUST)

jasonwestmas
09-03-2015, 02:17 PM
And there lies the problem. You need to have all the animation available to blend correctly




Except not all animation poses blend in a pleasing manner, which means you would have to predict perfectly ahead of time which poses work together and which ones do not. That's a very non-human and rigid way to deal with this problem. I don't want to have to bake things just to see if the poses from two different clips blend with each other. I suppose if there was an easy way to copy/paste/test between two poses before baking in the timeline. . . that would be a more flexible methodology.

lightscape
09-04-2015, 12:30 AM
Spinning logos was a figure of speech. :D
Baking is no a go workflow for me.