SIGGRAPH 2013 - Implicit Skinning: Real-Time Skin Deformation with Contact Modeling

[@makkon06]

I second the notion that we need this for games.

[@powergannon]

Can't wait for this level of dynamic deformation to be feasibly run in realtime

[@Teturarsyra]

At rest pose vertices save the value of the field-function (in other word their "distance" from the implicit surface). During the animation vertices seek their saved field-function value. This way we keep the same "distance" to the implicit suface even after deformation. As a result at rest pose they do not need to move to reach the original field-function value.

[@Teturarsyra]

External objects are not discussed in the paper. However this should be easy to handle if the model is skinned with our method. You can do rigid bodies simulation -> the character/external object will stay rigid without intersecting. If you want a soft deformation, you will need to approximate the external object with another field function and use a dedicated blending operator to merge it with the character. Deformation will then be handled automatically by the system.

[@MrPupps903]

Oh my GOD!!!! This so f***** cool. my most sincere congratulations!! Great work!

[@Evandroworks]

Nice work! The only (essential) thing I can see missing in your skinning solution is the increase of mesh volume on the perpendicular direction to the rotation axis (sideways to the joint/pivot point). For example, when you fold your arm or legs, the volume of your arm grows sideways on the joint (elbow/knee) because of tissue being repositioned there to accomodate for the self-collision. This is a specific type of volume preservation that is present in our bodies.

[@anishwij]

This can be done in cinema 4d using range mappers and pose morphs but it takes a lot of time to move the points to the correct position.

[@KnakuanaRka]

Great for animation and video games. Hats off to the guys who came up with this! Also, you misspelled partitioning as partitionNing at 2:28.

[@Teturarsyra]

Oops... And the font is pretty big as well! Unfortunately youtube won't allow me to re-upload the video without losing all stats and comments :(

[@lbarthe]

Codes under GPL 3 license available on the authors webpage (Rodolphe Vaillant and Loïc Barthe) from November 10 (2014).

[@Moebius3D]

Can we expect this to one day become a plugin for maya or another high end 3d animation package?

[@Teturarsyra]

As stated in the description: this is research work. There is no plugin available only a standalone software solely made to showcase the method.

[@armwulfpiloteer3255]

This shows clear improvement over existing methods in quality, but by what degree is the performance impacted?

[@Teturarsyra]

As you can see a character of about 4k vertices runs around 95fps. So it's not good enough for games (unless used partially maybe) but feasible movies

[@nvl901]

Would a baking approach work for realtime scenarios? Generating a deformation texture of some sort and having the vertexes transform according to the deform map or something like that.

[@Teturarsyra]

The 3D armadillo model comes from the Standford University scan repository I have personnaly never seen the original toy, it would be interesting to post a photo of it somewhere. The toy itself might come from a comic (c.f. wikipedia) or maybe it's just another warrior representation of the animal I don't really know...

[@Asdam12]

Holy SHIT. I know you guys aren't planning a plugin or anything but this would be very commercially viable, if your team could get its foot in the door. Educating people on KZbin is good, but you guys ought to be making hella stacks by this point!

[@FailingDarkness]

My second time watching this. :) This is such amazing work. This would save the rigger so much time.

[@Teturarsyra]

The software showed in the video is a standalone application developed by our research team. It has be designed solely to demonstrate our method and is not ready to be used by general public unlike the professionnal packages you are refering to. If our method is to be integrated in other softwares it will certainly take some time.

[@eliaslaurell5052]

We need this in games.

[@GameDevDave]

This could really speed things up for me... No more extra bones inbetween limbs for good crease-bends and it saves a bunch of weightpainting. :D when's this available? I subscribe!

[@Teturarsyra]

Indeed this can be a nice futur work. There is already some work done on volume preservation for meshes but it would be nice do some research on it with implicit surfaces.

[@Themindforest]

I hope that Autodesk considers hiring your team to integrate this approach into all their CGI animation packages. The current tools for dealing with joint deformations are extremely non intuitive and cumbersome. Your results are wonderful, but I also hope it is easy to implement and use if it gets into one of the animation packages someday.

[@JuicyVitaminStudio]

You actually addressed a issue, that really bothers me in games. Sure you can use custom blend shapes animations these days but your method is indeed the more elegant solution with great results. Hopefully some engine maker will take your papers and implement that.

[@robosergTV]

+JuicyVitaminStudio i am sure its all patented

[@Teturarsyra]

+Игровой канал RobosergTV Actually it is not.

[@robosergTV]

+Teturarsyra wow, didnt expect it. Is the majority of work presented on SIGGRAPH also free? Or does it vary?

[@Teturarsyra]

+Игровой канал RobosergTV It varies a lot. Some patent the algorithm, some only sell the code source or the software. Buying a patent is not always worth the trouble, sometimes just selling the software is enough. A lot of paper at SIGGRAPH are written by people hired in public institutions their primary goal is to share knowledge. Most paper aren't patented and provide free code source. For instance, take a look at this famous lab: igl.ethz.ch/publications/ you'll see a lot of SIGGRAPH papers with open source codes. Of course if you're private lab like Disney research the story is different...

[@robosergTV]

+Teturarsyra That is very good to hear. Thank you for the information.

[@Camtrack]

really nice , looks real ands looks easy to setup , Im really waiting to see it in 3dsmax

[@Themindforest]

Fantastic work! Those are some very impressive results! I would love to have such nice deformations on my models without doing a TON of tedious work that is required to emulate this effect by the current tools available.

[@IllSkillz]

2021 and why this is not yet implemented in to maya?

[@Teturarsyra]

For HRBF there is code on my homepage (see description). You'll also find how to contact me if you have specific questions.

[@EmDee3d]

This looks really great. Is your research team planning to intergrate this into 3D applications, like maya or softimage? Maya is a very open bit of software. I could see it doing well as a skinning plugin perhaps. Many studios would be happy to pay for the use of this.

[@anemoiapolisofficial]

Agreed; The Blender would love this. Unity too.

[@rodrod5150]

Hello Teturarsyra! I´m thinking to implement this paper on procedural ´Softimage ICE nodes´ to make a implicit skinning deformer. i found that HRBF is very hard to implement... any suggestion to make this step more easy? Links, explanations... hints? Thanks.

[@Taugeshtu]

All right, this is nice and good piece of work, but! While you do offer a comparsion in visual aspert with linear and dual quaternion methods, could you please provide perfomance data for those method on test scenes as well? At the end of a day, quality boost WITH perfomance lost is what goes into the decigion on whether to use a technology or not. If it's a fall from 300 FPS to 30 - that's not an appliable technology yet, but if it's 45 to 30 - that's a deal right there.

[@Teturarsyra]

Comparison is in the paper page 7 table 1 (rodolphe-vaillant.fr/pivotx/templates/projects/implicit_skinning/implicit_skinning.pdf). Yes the performance drops dramatically compared to these methods, however, I cannot agree or disagree with your statment because everything depends on context and target application. Obviously you won't use Implicit Skinning to fully skin every characters in a game, but you may use it for some close up shots (a hand for instance). If it's for an animated movie, being able to interact with a single character can be enough as the rendering is always done offline.

[@Taugeshtu]

Teturarsyra Ah, I missed the paper link :D Well, hands down, with its own limitations, this technology is appliable in interactive scenarios :)

[@Xitement]

such a nice in between result. really close to where it should be. really quite amazing for something automated like that. good job makers! Is this the official youtube channel for these makers? And where can we support or contact them.

[@mishominas]

why you don't offer your product to Autodesk Maya ?

[@CGFUN829]

this is a method not a product. he will implement these methods in all 3d apps.

[@kebrus]

there's a part of the technique i don't it, when the vertices are projected onto the field functions how can you define vertices going away from the joint with gradients? and how do you handle these projections when there's no changes to the bones, lets say the bones are in it's rest pose, why aren't vertices trying to match the field functions?

[@Zoza15]

2:01 I don't see the muscles react to the bend of her elbow.. If you try to correct the bone structure, then also correct the muscle structure as well. Despite the little improvements, you guys are still a long way improving the entire meshed body..

[@Teturarsyra]

That's because we don't simulate the muscles of course. The technique "only" handle automatically these: skin contacts and transitional effects at joints (bulge , blend). Muscles are yet something to explore, but I'm afraid it'll always come at the price of more user edition. Don't forget also that all results here are produced by a single researcher who has never been trained to use CGs software. Math and programming is our daily lives, the rest is done once every six months for conference demos. But I hear you, there's definitely room for improvements. And that's great because it means we still have plenty of cool things to work on :)

[@krisCrashTV]

I had super good results combining Linear and Dual Quaternion, painting a mask of where to use one or the other. I don't know if this is supported by games though ;o

[@ochgottnochma]

You really should offer some SDK with demo programs... then will ppl dig up with animations, embed that into games, contribute code back etc. (if license will allow it) + GitHub

[@Teturarsyra]

No worries It'll come. We're hoping to release the prototype under open source license sometime before this summer.

[@hermannrolfes7744]

Teturarsyra Thats great to read, thanks! :)

[@MKDAWUSS]

And what happens when a foreign object presses against the model? Like clutching a ball, or leaning on a wall?

[@nickolasoddo6540]

what is the armadillo model i have a old toy of that model dated 92 or 96 i always wandered who it was

[@hadriscus]

How does your method fare in terms of performance compared to linear and DQ ? it's a rather important aspect, unfortunately your comparison shots at the end don't mention the FPS from the classic methods.

[@Teturarsyra]

Comparaison is described in the paper (see description). For instance the hand runs at 95fps with DQ against 15fps for Implicit Skinning. Obviously with new hardware the number would change, also you could possibly make IS faster by only applying it to a subpart of the mesh vs running the deformer on the whole mesh.

[@waronmars]

Would this sort of skinning be practical for realtime applications like video games?

[@kingdavidjapan]

So when will we see this in an application? Never, right?

[@adamnixon84]

Was this done only in Blender or was it also done in other 3D packages? I'm mostly interested in Maya but would be interested if it were also tested in 3dsMax, C4D, or MotionBuilder.

[@Henglaar]

There's a lot of programs that would benefit from this. As I use Poser, I have to agree with KleanZlate that this would be a great new feature for Poser. I'm certain that Lightwave, Maya, and other users would be equally happy to gain this. How can I help make this happen?

[@AarmondHybridDixon]

anyone already compile this for windows? im stuck and cant figure it out....

[@Gauntlets]

Now if only there was an actual tutorial somewhere 4 both MAX (preferably 4 me) an' Maya...

[@beingatliberty]

nice work well done a real improvement

[@maayaa77]

isnt maya muscle used for real world skin contact distribution.

[@TheXrabidx]

Is this skinning method robust and viable for real time applications?

[@Teturarsyra]

Mmh, indeed we can extend the length of our bulge. However it acts like a "wave front" following the intersection of the two limbs, I'm not sure it would be realistic for a bicep. It might not be so hard to extend our method for biceps bulge, but it hasn't been investigated yet.

[@Teturarsyra]

There is no plan to do so in our team. But we will gladly help anyone who wants to do it.

[@alexlei2235]

Once graphics get to their full potential, they start do holo-gaming. Once they reach that, they do single-player real-time virtual world gaming. Finally, online, real-time, mindblowing physics, realistic graphics, multi-region/category, virtual world.

[@KleanZlate]

Please contact Smith Micro and get this into their Poser application :)

[@krissosful]

Amazing job! Keep going:)

[@alef321]

fuck this ,we had zero animation tech and real gamers loved it ,just look at MML1 or classic Resident evils!! ok now that everyone got mad I gotta say, this is amazing

[@JoshPurple]

This looks GREAT :) ! Thank you!

[@a3rodox]

is this going to be a maya plugin?

[@UpsideDownMon]

great stuff, it looks a lot better

[@landinboyd5573]

It's 2017 can this be used for games now?

[@IvanTsyvinskiy-u1q]

10 years AGO!!???

[@Teturarsyra]

I know 😭

[@zechordlord]

Well done, impressive method!

[@RobertDigitalArtist]

Sweet, where do you get the standalone software??

[@Teturarsyra]

Take look here for the research codes: rodolphe-vaillant.fr/permalinks/implicit_skinning_project.php#CODE

[@RobertDigitalArtist]

Teturarsyra Great, appreciated! Is it stand-alone, how do I use it?

[@Teturarsyra]

Robert LC - Digital Artist Standalone and NVIDIA only. You have to compile it. We provide the sofware and codes only as a mean to test and reproduce our work for other developers; it was never intended to be used by professionnal artists in production pipeline. Our hope is to inspire developers to code more viable, robust and user friendly plugins.

[@kindkind4813]

plugin maya?

[@ambidexter2017]

Knees still look unreal. I suspect the problem is that most human models don't really take the kneecap into account…

[@Teturarsyra]

Yes, that's true. A solution would be to allow the artist to model the kneecap as an additional implicit surface shape. Downside is it will take a little bit more time to setup the character.

[@sanketvaria9734]

just tell us how to do this in maya

[@PashaDefragzor]

3:55 cool heavy cinematic hand

[@gsmadmax100]

where can i get this ?

[@JarvidO]

It's safe to assume this isn't open source

[@lbarthe]

If you want to see the recent advances around the implicit skinning, follow us on Facebook : facebook.com/implicit.skinning/

[@alexanderleeart]

beautiful!

[@theicychameleon]

Thats really impressive!

[@Teturarsyra]

For the moment no. We're still too slow to pretend to be used in such applications, games requires thousands of effects in addition to skinning. Implicit Skinning would be more appropriate in animated movies.

[@h3v3ns3nt1]

i need it for max

[@FriedRiceBoy]

oh my god!! thank you!!!

[@madnessinmotion]

Amazing!

[@DonOmQ]

does this work in 3d max and character studio? thank u

[@dasten123]

amazing!!!

[@blackrack2008]

awesome

[@Evandroworks]

Here is an example of what I'm talking about: watch?feature=player_detailpage&v=spIRr9OcODk#t=67s

[@tummibearx5025]

man if only bio ware knew about this.

[@tedeusz83]

that looks great, i hate skinning and its problems :)

[@sione_etc]

oh man imagine this + euphoria engine

[@Nulono]

*partitioning

[@1r0zz]

I do not want to be too fussy, but it's very rubbery and not realistic. it is an improvement, and I honestly think it's a bloody good work, but it's still far to be realistic...

[@Teturarsyra]

Well we never aimed for photo-realistic results, so no offense taken. In this work we try to improve the look while keeping real-time to interactive speed. So if you like the improvement we reached our goal :)

[@1r0zz]

Teturarsyra I am a profane to 3d animations, and I am just flapping my mouth. for make it more "realistic" could not you use some kind of animated normal maps?

[@Ramt33n]

I suppose it's really up to the artists to use these tools as ways to produce realistic or stylized results!

[@Teturarsyra]

1r0zz Well yes, we only focus on the deformation of the mesh. Any subsequent effects will make it look better of course.

[@1r0zz]

Ramteen sure.

[@gahjii]

i looked at my finger if it bulged

[@sleepCircle]

hot dang

[@Teturarsyra]

Never say never :)

[@lbarthe]

Improved version under commercial development. See here for more information : www.irit.fr/~Loic.Barthe/transfer.php

[@anemoiapolisofficial]

*Blender Community

[@romfromrussia4211]

too slow yet