PDA

View Full Version : Revamping lightwave character animation



lightscape
01-24-2015, 04:35 AM
With genoma and rhiggit helping lightwave to progress further. The limits of the old ca system in lw is just not cutting it in a fast pace environment. I couldn't work in lightwave if not for Rhiggit.
How can it be automated? Rigging, skinning, mocap, converting to a control rig. How fast can it possibly be?


https://www.youtube.com/watch?v=nEKBqf1Ze3M&list=WL&index=18




We are doing this with digital doubles and cut the time to do it by 80% using maya and 3dmax biped.
Newtek should license Human Ik from autodesk.
Make a robust system around Human Ik.
Partner with Mixamo that has the best autorigging system which uses Human Ik as a base.
Stop working with antiquated skelegons in modeller. Officially ditch Nevron since kinect 1 is dead.

No I'm not paid by Mixamo. We just use it and love it.

Marander
01-24-2015, 08:47 PM
This is is crazy rigging:

http://quantum-human.com/

http://vimeo.com/113461164

- - - Updated - - -

This is is crazy rigging:

http://quantum-human.com/

http://vimeo.com/113461164

lightscape
01-24-2015, 10:25 PM
Yep saw that in the foundry forums. Its in beta.

Btw Mixamo has a way to create morphmaps automatically.

http://s30.postimg.org/f3r0lg3lt/morph.jpg (http://postimage.org/)




https://www.youtube.com/watch?v=e_RnNYBtA3Q&index=21&list=WL
Newtek...partnership...develop them with 3dprinting companies, mixamo, ikinema, autodesk.
Licensing human ik like what mixamo, EA games, etc, did could speed up character pipeline in lightwave.
We need more speedy workflow for previz, digital doubles, etc.

lightscape
01-24-2015, 11:15 PM
http://usa.autodesk.com/adsk/servlet/item?siteID=123112&id=14255040

Why did such a capable team decide to integrate HumanIK instead of writing their own animation middleware? “From the moment we got HumanIK to the moment we were first able to run it on our pipeline was approximately one week. This is very fast in terms of integration. The product is very modular and can be plugged easily into existing pipelines.”

“If you were to write your own IK system, you would develop for 6 months to 1 year and wouldn’t really be sure if it met the quality bar until that initial, heavy investment,”


:newtek:
Mixamo is not the only dcc app that licensed HumanIk from autodesk afaik.
Modo meanwhijle licensed Ikinema instead of making their own. http://www.ikinema.com/?mod=press_release&show=15

Mixamo has the edge so far with autorig, autoweight, automorph and well integrated into many appz and the autoconvet to max and maya rig is unbeatable and fast.

magiclight
01-25-2015, 01:19 AM
I do not understand the comparison, if you compare features 3DS/Maya vs LightWave you need to compare the price tags also.

Licensing any AutoDesk tech is not cheap, don't think for a second that it would not have a huge impact on the price of lightwave,
I have never used Mixamo, but this is a subscription solution that fit many people and the kind of stuff autodesk love, but all people
are not happy having all their data in a cloud solution controlled buy someone else.

(I am sure lots of users don't care of the price of LW but I am also sure that a lot of users do care about the price)

I am sure 3DS/Maya has lots of cool features missing from LW, but they have a price tag that is missing from LW also.

New cool features are always nice but it takes time to get them implemented and it all depends on how much resources the
development team have access to, when it comes to licenseing other tech it's not that easy, it has to fit into the rest
of the system, you cannot just plugin some other tech into LightWave and snap your fingers and expect it to work,
just because something is easy to integrate into one piece of software does not mean it is easy in another,
there are lots of license questions to work out and so on, it's not just "lets buy this tech and plugit in to our software".

Reality is far away from that simple.

lightscape
01-25-2015, 05:21 AM
Just take my word for it. Its not expensive to license human ik. But the benefits of having it is immense. just search the practical usage of human ik on the net.
To develop your own character animation tool is also more expensive and time consuming and like the article says a heavy investment.
Also note Mixamo is a lot smaller than Newtek yet they were able to afford to license human ik.
Luxology was also a lot smaller when they integrated ikinema. Brad sees potential in some technology that newtek should have, too. Groboto was first seen in newtek forums but largely ignored. Now its well integrated into Modo.

jasonwestmas
01-25-2015, 09:16 AM
well tbh I really don't find rigging and autorigging options to be Lightwave's weakest link. More like deformation control based on the rig. Also the animation keyframing process is in great need of workflow improvement as well. So far I'm impressed with what has happened with the bullet dynamics setup in lw2015. Very direct way to set things up and make it work quickly. I hope the same kind of "directness" will happen with the keyframing, saving poses, making selection sets and the animation mixing tools. As far as deformations go LW is greatly lacking and I can't imagine they would let that go for much longer.

Also I was not impressed with modo's implementation of Ikinema in the sense that I could not even do basic things like mirror joints correctly in modo 601 or 801, where as in all other applications it's a snap. Just seems kinda sloppy over there.

lightscape
01-25-2015, 09:47 PM
well tbh I really don't find rigging and autorigging options to be Lightwave's weakest link. More like deformation control based on the rig. Also the animation keyframing process is in great need of workflow improvement as well. So far I'm impressed with what has happened with the bullet dynamics setup in lw2015. Very direct way to set things up and make it work quickly. I hope the same kind of "directness" will happen with the keyframing, saving poses, making selection sets and the animation mixing tools. As far as deformations go LW is greatly lacking and I can't imagine they would let that go for much longer.

Also I was not impressed with modo's implementation of Ikinema in the sense that I could not even do basic things like mirror joints correctly in modo 601 or 801, where as in all other applications it's a snap. Just seems kinda sloppy over there.

That is the beauty of having a standard system. Its easier to create templates, save poses, blend motions, animation layers because there's a system already in place for that. Everything that lightwave or modo can do now, other 3d apps can do years ago and more.
We don't have animation layers, no clear way to identify an entity(character), etc. There's no standard system in lightwave to adhere to. So support from external apps is always a problem. Without rhiggit, transferring rigs would have been a huge headache.
Motionbuilder has the most complete toolset and human ik is derived from it. Its robust and tested. Autodesk is licensing it. Its supported by a lot of external apps. Even smaller devs are licensing it.

Regarding bullet, its even easier in other appz and they didn't need to have a paid update to implement motors, hinges, etc.

Mr Rid
01-26-2015, 01:31 AM
This is is crazy rigging:

http://quantum-human.com/

For many years, that's the only rigging tool I have been interested in - one button. After zillions of human figure rigs made over decades, its time for it to be a few clicks. I don't like rigging. Its always a pain. No auto-rig tool is actually "auto" if I have to paint weights. Weights are tedious, inefficient, stuck in the mesh, they make or break auto-rig deforms, and LW weight tools leave something to be desired. It does not make sense to have to monkey in Modeler in order to control bone influence in Layout. Auto-rig tools inevitably depend on you painting weights, they somehow never work on my mesh at default settings, and I still have to become an expert in rigging in order to edit the rig to fit, without breaking it.

I really think weights could be bypassed in many cases by updating the limited range bones in Layout, with GL click-n-drag shapes and falloffs.

A.R.T. looks wonderful if ever released- http://www.anzovin.com/art/ excellent automatic deforms without weight painting, but the weight painting is also very simple and intuitive.

RebelHill
01-26-2015, 06:03 AM
I think you're slightly misunderstanding what humanIK is... it's just a solver, nothing more. Whereas we're used to seeing things like a 2bone ik system, or 3 bone, whatever... humanIK is simply a full body IK solver, that's it.

I don't believe that mixamo do licence it at all, simply, they provide a bone setup in both their rigs and animations, which humanIK understands (this is basically the good ol standard fbx hierarchy/naming convention). Make the same hierarchy in LW and it'll drop straight onto humanIK in maya (or even without you can reassign the specific, recognised parts). Mixamo don't offer (so far as I have seen) any kind of manual animation tools, thus having humanIK would be pointless... all they need to do is retarget mocap from one body to another, and believe me, writing a retargeter is hardly any kind of effort (took me all of a day to do).

The only reason it's easy to copy motions/poses/etc using it is because there is only the one rig... with no variation. This is no different in LW... use the exact same rig for every character and you can copy stuff back n forth no problem (retargeting notwithstanding).

This kind of "single rig standard" though isn't the full story, as it really is just the one rig and character type you have available. Sure human like characters are the most common requriement, but in all my years, thats included only about 2/3 of what folk've wanted or needed. The other 1/3 has been non standard human stuff, or other custom need. Same thing with that quantum human... only good for standard human meshes... 2 arms/legs... 5 fingers, etc. The only way to do this stuff automatically is to have a fixed body model, and to conform all inputs to it.


Auto-rig tools inevitably depend on you painting weights, they somehow never work on my mesh at default settings, and I still have to become an expert in rigging in order to edit the rig to fit, without breaking it.

You should try RHiggit... weights are easy to deal with and commonly require about 10mins for some basic maps.

Mr Rid
01-26-2015, 07:06 AM
You should try RHiggit... weights are easy to deal with and commonly require about 10mins for some basic maps.

I tried RHiggit Free v1 awhile back and I ran into the problems I mentioned, and was unsure how to get it to work. I am sure V2 is better, whenever I get time to try the demo.

RebelHill
01-26-2015, 07:23 AM
I tried RHiggit Free v1 awhile back and I ran into the problems I mentioned, and was unsure how to get it to work.

I can barely imagine how... and you couldve asked.


I am sure V2 is better, whenever I get time to try the demo.

There isnt one Im afraid.

Mr Rid
01-26-2015, 07:54 AM
I can barely imagine how... and you couldve asked.

I'll email you. At the time it was not crucial, and I felt it was not worth messing with if I had to ask.


There isnt one Im afraid.

What is "RHiggitFreeV205.zip" on your site?

Back in LW 7 days there was an autorigger that worked decent, and it auto-painted weights so you could swap meshes easily. I dont remember the name, but as I recall it disappeared after the author had sold it to some company that promptly squashed it, and forbade the author (who was very upset about it) to talk about it.

ACS4 was a complete waste of money. I could not get it to work and Pazera was not helpful.

RebelHill
01-26-2015, 07:58 AM
Thats the free edition... contins cut down versions of the rigging and animation toolboxes (which in themselves give you some much needed tools in LW, or streamline certain procedures to make common tasks a ton faster)... but theres no bodybuilder in there. Painting weight is SO easy though... just block out sections to prevent cross influence (like between fingers, or the legs) and that's pretty much it. There's also a tool to come that lets you load the bones back into modeler as skeles, so you can autoassign from vertex paint or the like... but those autoweights always need cleanup.

If you could autoweight that perfectly and easily... there'd be no need for weights at all.


https://www.youtube.com/watch?v=aXj8NIiVyxs

lightscape
01-26-2015, 08:14 AM
I think you're slightly misunderstanding what humanIK is... it's just a solver,
I don't believe that mixamo do licence it at all, Mixamo don't offer (so far as I have seen) any kind of manual animation tools

The only reason it's easy to copy motions/poses/etc using it is because there is only the one rig... with no variation. This is no different in LW... use the exact same rig for every character and you can copy stuff back n forth no problem (retargeting notwithstanding).

This kind of "single rig standard" though isn't the full story, as it really is just the one rig and character type you have available.
The only way to do this stuff automatically is to have a fixed body model, and to conform all inputs to it.





Its a system to start something like characterstudio. They did say in the articles that it would take a long time with heavy investment to develop it themselves. Both mixamo and lux, EA and probably some more companies, they took what's already existing and licensed it and build on it. I think we will see modo with a better animation system in 901 if they survive the sale of the Foundry.
Iclone also licensed human IK and the animation system allows manual animation tools.
You're right its easy to copy, pose, etc because it is only one standard system. Covering 2/3 of peoples need for a standard system is better than 0 system imho. Characterstudio might be a human biped system but its not hard to make it into a a horse rig, a dog ,etc.
We don't have any standard in lightwave. There is no fixed system to adhere to, to conform inputs into. Rhiggit is thirdparty and not everyone has it unfortunately.

Genoma and rhiggit are attacking the problem in different directions but both not communicating at all to create some sort of standard system. So external apps always have no templates for lightwave.

lightscape
01-26-2015, 08:19 AM
You should try RHiggit... weights are easy to deal with and commonly require about 10mins for some basic maps.

Creating weights is easy in lightwave indeed but getting the deformation 80% there takes time. That's probably what people spend the most time with.

RebelHill
01-26-2015, 08:32 AM
Luxology DONT use HIK... they use ikinema (which in many respects is superior). But again... HIK isn't any kind of system in the way you're describing (and it's been around a long while now)... It's just a full body solver, thats it. A sort of fusion between a standardised, preset rig, and the IK/FK control structures placed into it. You copy from one to another... because the parts/controls are the same set on every rig (cos there is only one rig)... retargeting, thats another matter, but again, facilitated by the fact that there's only 1 rig.

I agree that some sort of standard is system is beneficial... that's how RHiggit works underneath. The parts are standardised and can be compared to one another, that's why its possible to take poses or motions from a human character, and paste them onto a dinosaur. The 2 rigs can be very different, but can be cross interpreted between one another. Genoma seems to think that building a rig alone is enough, and animation is an afterthought at best, rather than being an ingrained aim of what its trying to accomplish.

lightscape
01-26-2015, 09:08 AM
Luxology DONT use HIK... they use ikinema


Ehem. I did say that mister. :D Post 4 And its in one of the article I posted there.
IKinema’s technology enables animators to rig characters and retarget motion capture data in a fraction of the time and cost of alternative solutions.
http://www.ikinema.com/?mod=press_release&show=15

Either system newtek adapts is ok as long as we get a standard system that is already adapted by external appz. Newtek is creating its own island again with genoma.


Anyway a gift for you in pm.:D

lightscape
01-26-2015, 09:36 AM
Here's the Rhiggit human model I uploaded and rigged in 1 minutes 15 secs.

http://tinypic.com/r/2vnkz7a/8

I can import this in max or maya and convert to a standard biped rig that exists in both app's standard animation system retaining all the autoweights, too, but with controllers and all. That's 80% work there in less than 2 minutes.
I know for sure Rebelhill will have the same scripts soon :D but those that don't have rhiggit won't be able to enjoy the time savings.

RebelHill
01-26-2015, 09:47 AM
"Our instant algorithm takes up to 2 minutes"...

Not really sure what to make of that sentence.

jasonwestmas
01-26-2015, 11:10 AM
Yeah I think what I'm really going on about in native lightwave keyframing is the lack of standardization TOOLS and GUI and the ability to lay down keys to edit them in a more predictable/manageable manner. RHiggit shows this kind of thing off nicely with its parts system and the cool curve mode buttons all tucked into same tool window. I just wish it was easy for me to create my own parts system on a whim like I can in maya using the character sets stuff. I don't want to have to script my own tools, which I guess is my main point. heh.

Editing vmaps directly on a rigged character in a pose would be my number2 hope. It's so much easier to do minimize clipping with characters that are carrying gear and getting those shoulders waist deformations looking slick in more extreme poses. It's important.

Mr Rid
01-26-2015, 09:23 PM
... If you could autoweight that perfectly and easily... there'd be no need for weights at all.



Seriously, someone needs to think in a new way that bypasses weights. Forget weights! You should never be in Modeler to control bones. That has never made sense. Bone influence should be controllable in the bone properties, and should have nothing to do with geometry. Offhand, I picture a cage around the bone that shows in GL that you click and drag on, to shape it.

jasonwestmas
01-26-2015, 10:12 PM
If geometry has nothing to do with bones then how would you calculate the deformations based on bone/joint rotation angles? Sounds like what you are describing Mr.Rid is a cage deformer. . . but a cage deformer still needs to rotate with the skeleton some how.

lightscape
01-26-2015, 10:57 PM
"Our instant algorithm takes up to 2 minutes"...

Not really sure what to make of that sentence.

Nothing malicious in the background I hope :thumbsup:


Seriously, someone needs to think in a new way that bypasses weights. Forget weights! You should never be in Modeler to control bones. That has never made sense. Bone influence should be controllable in the bone properties, and should have nothing to do with geometry. Offhand, I picture a cage around the bone that shows in GL that you click and drag on, to shape it.


Envelopes in 3dmax are more developed but you still need to go down to the vertex level to adjust weights for adequate deformation.
Its really time consuming in 3dmax and lightwave has it easier with the bones having pretty good bone falloffs.
But I don't know how we can get rid of weights. How will a vertex on a geometry know where to move to if there's no weights to tell it?
A cage deformer like 3rdpowers has less precise control on deformations. Its still useful though.

tonyrizo2003
01-26-2015, 11:41 PM
Seriously, someone needs to think in a new way that bypasses weights. Forget weights! You should never be in Modeler to control bones. That has never made sense. Bone influence should be controllable in the bone properties, and should have nothing to do with geometry. Offhand, I picture a cage around the bone that shows in GL that you click and drag on, to shape it.

I disagree with you about weight mapping, we have tools now that enable you to easily weight map anything you like. Not to mention that maybe you're not utilizing weights in the way they are meant to be used? I know that sounds like a knock but it's not meant to be. Many times we users think a tool should work a certain way, and when it doesn't work the way we think it should, then we get mad at the developers. However, once we find out the parameters for that tool and work within its limits well that's a different story. An example would be IKB and what Ryan Roye has done with it, as well as the rigging system that RH has created, all within layout.

You don't have to use weights. Larry (Spline God) didn't always use weights to animate characters, he used the bone falloff tool, I suspect that people wanted weights because its a more visual way of setting limits.

You can also use vertex paint in modeler, once more with vertex paint you can also adjust the rotation of the bank handle for the bone as well as see the deformation and adjust the weight map on the fly all within vertext paint.

Weight painting in Layout.

http://forums.newtek.com/showthread.php?117565-Weight-Paint-Tool-in-Layout
https://www.youtube.com/watch?v=OiBhni1wzuU
http://www.3dcel.jp/

I've seen this but never played with it as maybe Dennis has, this might be a solution you are looking for.

I know you are a non programmer such as myself, there is nothing preventing you from finding a programmer that might be able to write such a tool for you specifically that you can own out right. Especially now that LW has Python there are quite a lot of Python programmers out there. I realize that you would like LW to have the best solution so you don't have to spend more money and I think a lot of us would. It's just not the reality, if LW3DG did that than LW might have been nothing more than vaporware. That's one of the beauties of say Maya and 3DS Max, they give you the ability to write your own tools if needed. I'll give you a great example of Timothy Albee, he didn't write TAFA Mac Reiter did, http://ta-animation.com/FA/

Just my two cents.

lightscape
01-27-2015, 12:05 AM
You don't have to use weights.

In some cases yes. But in some cases no. It would be a rubbery affair without proper weights. Hold bones do work but sometimes when working with other apps it becomes problematic especially with lightwave having reversed weights using joints.

tonyrizo2003
01-27-2015, 02:58 AM
Lightscape, you're absolutely right, it depends on how much control over your mesh you need in LW, as for other apps I have no clue as my main 3D package is LW.

jasonwestmas
01-27-2015, 07:20 AM
Lightscape, you're absolutely right, it depends on how much control over your mesh you need in LW, as for other apps I have no clue as my main 3D package is LW.

In other apps. it's sometimes a case of having in depth weight mapping tools that can be used on the model in any pose you wish, which is what I prefer to make sure that extreme poses look the best they can be. They don't have bone falloff and field weighting but it really isn't necessary since the field "weighting" is handled when the skin weights are first applied and then you take control from there, making the most appealing modifications.

tischbein3
01-27-2015, 08:40 AM
Seriously, someone needs to think in a new way that bypasses weights. Forget weights! You should never be in Modeler to control bones. That has never made sense. Bone influence should be controllable in the bone properties, and should have nothing to do with geometry. Offhand, I picture a cage around the bone that shows in GL that you click and drag on, to shape it.
I really do think these things should be rather be weight map modifiers wich do work on top of a weight map ( and are bakeable to the maps). Weight maps are good, but they really need to be pushed further towards useabillity in setting those up.
This also includes current weight map baking (wich includes vertex paint and my experimental, myself not beeing satified with the result, plugin found here http://www.splotchdog.com/index.php?option=com_content&view=article&id=58:cuv&catid=11:lightwaveplugins&Itemid=161 ) might need to be enhanced to use some "simple" volumetric mapping approaches
http://blog.wolfire.com/2009/11/volumetric-heat-diffusion-skinning/
(Still doesn't have made up how it should behave on non closed meshes)

tonyrizo2003
01-27-2015, 09:20 AM
I really do think these things should be rather be weight map modifiers wich do work on top of a weight map ( and are bakeable to the maps). Weight maps are good, but they really need to be pushed further towards useabillity in setting those up.
This also includes current weight map baking (wich includes vertex paint and my experimental, myself not beeing satified with the result, plugin found here http://www.splotchdog.com/index.php?option=com_content&view=article&id=58:cuv&catid=11:lightwaveplugins&Itemid=161 ) might need to be enhanced to use some "simple" volumetric mapping approaches
http://blog.wolfire.com/2009/11/volumetric-heat-diffusion-skinning/
(Still doesn't have made up how it should behave on non closed meshes)

@tishbein3, I recently tried using skeleweights but it didn't work for me, not sure what I did wrong as I followed or at least thought I followed the instructions on how to use it with Genoma. Any ideas? Thanks :)

This system is pretty sick though!!
https://www.youtube.com/watch?v=-RpqbC5-Z0E
https://www.youtube.com/watch?v=nPUxnm5qscA

tischbein3
01-27-2015, 09:25 AM
care to send me the file ?

edit:
One possible reason it does not work is, that the distance to the bone is to high, in that case you have to _lower_ the scale factor,
for having a softer falloff.

tonyrizo2003
01-27-2015, 11:22 AM
Ok thanks, I'll have to do it when I get home from work later this evening. Thank you :)