c4D quadruped skinning

[sonic-blade]

Hi Mark!

I think (can´t know without seeng it) your mesh might be too low-poly around the shoulders.

If you want to go the third-party-software-route you should take a look at motion builder.

jm2c
Volker

[JamesMK]

There's simply no way you could away with just autoweighting on something like a jaguar (or any other mildly complex multi-limbed organic thing for that matter). Keep in mind that the construction of the mesh is highly important here too - it's quite easy to build a model that's close to impossible to deform correctly, and no amount of weighting will save it.

Given a good mesh, the only way is to get in there and do the weight painting until everything deforms smoothly, add corrective morphs and/or local deformers where needed and so on. There are no real shortcuts unfortunately.

And it would be the same in Maya.

EDIT: Besides, any big cat would be quite a challenge with the huge range between the extremes in their stride. So, it's merely a matter of investing lots and lots of time in getting it right. /EDIT

[shibunkin]

Thanks all for responding, I've attached a JPEG of the mesh with HyperNurbs switched on. You may be ale to tell me if I'm using too many / too few polys.

I was also keen to investigate how to apply the corrective morphs and local deformers mentioned by James MK - could you tell me what tools I would use for that so I can look into that.

Thanks

Mark

[soccerrprp]

Take a look at this thread and perhaps contacting Brian would be helpful.

http://forums.cgsociety.org/showthread.php?f=47&t=700279&highlight=bear

Check out his bear video. Is there a pre-made quad rig in the content manager?

Anyway, C4D certainly has the tools to do it. Just a matter of understanding how to use MOCCA or CDIKTools and understanding how quad motion behaves...


Richard

[Kokosing]

I think you may have to get down to some per-vertex hand weighting. Also, don't forget you can use corrective shapes as a last resort. By this I mean morph-targets which correct bad deformations. These can be driven by joint rotation so they occur automatically.

If it's any consolation I had a hell of a time trying to weight a dog when I was first learning to use the new Mocca. One thing I would suggest is looking carefully at some reference video. I was pushing my model to impossible extremes and getting annoyed when I couldn't make the weighting work.

Good luck,

W

[shibunkin]

thanks for the tips Will,
Could you tell me what you mean by corrective shapes and morph targets? Do you mean using the morph tag and adjusting the polymesh for the extreme positions when the mesh breaks down?

The big unknown for me is: is my mesh too complex or too simple? I guess I want to be sure the mesh in 'ballpark' before spending the time manually painting weights on it - is that what you mean by per-vertex hand weighting?

I've got some good ref video from getty and BBC motion gallery, thanks. My joints look pretty natural, I can't see them going beyond the natural limits of the big cat run cycle.

Cheers

Mark

[sonic-blade]

So far the mesh looks reasonable to me, could you post a screenshot showing your joints and the HN-cage?

[Kokosing]

Well, first off I should say that you'll get some much more intelligent advice from some of the more rig-savvy users on here.

But, yes, that is what I meant by corrective shapes. Get your weighting 90 percent right. After that you can iron out small problems with a morph. So if you're using MOCCA you can assign a morph target to the mesh and reposition vertices that don't behave. Using the range mapper node in Xpresso you can connect the morph target to the relevant joint's rotation. This way the morph target is only active when the joint is rotated. (There is actually a way in the morph target tag to associate the target with a joint and not get into Xpresso, but I haven't played with it much.)

Do to a CGTalk problem that makes me want to put my head through a wall, I can't see your jpeg.

You can troubleshoot your mesh by asking a few questions: what are the vertices doing when the joints deform the mesh? Is their layout (topography) such that they can move in a logical way, maintaining volume and form, when the mesh bends?

I found that quadrupeds can be tricky in the shoulder area because what's happening to the mesh isn't that obvious. While the skeleton may be moving around loads, the skin deformation is subtler and not always in the same direction as the joints. So you may find it easier if you don't rely on the joint movement to directly drive all the skin/muscle shapes. Cinema is full of neat ways to create muscle movements etc. Just check out the files that come with MOCCA.

Good luck, Will.

[shibunkin]

Hi Sonic-Blade, I've uploaded a JPEG of the joints - you'll see there are ribs in there to try and hold the stomach in place, and some pelvis and shoulder blade joints to try and get that distinctive movement under the skin around the hips and shoulders.

thanks

Mark

[shibunkin]

Will, that's super useful.

Last questions before I go away and do the r&d:

other methods of creating muscle movements in MOCCA - do you mean the CLUSTER and MUSCLE deformers?

By files that come with MOCCA - do you mean the FLABIO tutorials and the WIPIX rabbit?

ta


Mark

[Cactus Dan]

Howdy,

Well, I have 3 comments on your mesh:
Edgeloops, edgeloops, edgeloops. :D

But seriously, the most important things to get good deformation are the mesh's topology, skin weighting and the placement of the joints. A lot of deformation problems can be fixed by simply adjusting the topology of the mesh or adjusting the position of the joints. If the mesh doesn't have the proper edgeloops in it, it is going to make your work more difficult. Corrective rigging with morphs or clusters or any other methods should be used after all else fails.

Your original mesh (HN off) looks pretty good, but you may want to add a few loops in some key areas where you're having trouble. Sometimes just moving the original position of the points can dramatically change the deformation.

Here is an example:
The original downloaded mesh:
http://www.cactus3d.com/OriginalMesh.jpg

... and the mesh with the points adjusted:
http://www.cactus3d.com/TweakedMesh.jpg

The adjustment was made after the mesh was already skinned, so it was not re-weighted. You'll notice a big improvement in the deformation after the points were moved.

Another common mistake a lot of people make when rigging a quadruped is that they place the spine joint chain up at the top of the model's back. I know that's where an animal's spine really is in real life, but it causes deformation problems when skinning a 3d model. The spine joints should be placed closer to the center of the mesh on quadrupeds, to get better deformation. By placing them in the center of the mesh, there's no need for additional rib joints, which usually don't help all that much anyway.

Notice the placement of the spine joints:
http://www.cactus3d.com/HorseJoints.jpg

For the rig in the above examples, no corrective morphs were used. Only edgeloop and weight adjustments were made to improve the deformations of the original downloaded mesh.

And the final result:
http://www.cactus3d.com/Pegasus.mov

Getting good deformations through adjusting the edgeloops and weights is tedious work, but it pays off in the end.

I hope that helps.

Adios,
Cactus Dan

[shibunkin]

Hi Dan, thanks for your detailed reply. That all sound good advice, I'll to get the mesh as good as possible before doing corrections with morph tags and the like.

One question - in the pegasus example I noticed some triangles. Is it ok to triangulate a quad to add more points in the mesh?

Cheers

Mark

[Cactus Dan]

Howdy,

Triangles aren't always bad, the main reason to avoid triangles is that they don't really subdivide well when you put the mesh in a HyperNURBS. That particular model had those triangles in it originally when it was downloaded, and since the mesh already had enough subdivisions and wasn't going to be used in a HyperNURBS, I didn't worry about them. If it was going to be subdivided further, then I would've edited the mesh to eliminate as many of the triangles as possible. ;)

Adios,
Cactus Dan

[Horganovski]

Very nice looking rig Dan, for the neck would use regular Spline IK or a CDSpinal IK?
I'm building a cartoony horse rig for my own practice and not sure how best to approach that area.

Cheers,
Brian

[Cactus Dan]

Howdy,
Very nice looking rig Dan, for the neck would use regular Spline IK or a CDSpinal IK?
I'm building a cartoony horse rig for my own practice and not sure how best to approach that area.
For that rig I used a CD Spinal tag, but either one would work. ;)

Another thing to consider when modeling a quadruped is the base pose of the neck. Most of the quadruped models I've seen have the head and neck modeled in an upright position (probably because they're thinking "biped" when modeling). But, that is the upper extreme pose for the head and neck, and binding the skin in that position is going to cause deformation problems when you want to pose the quadruped's head and neck in the other extreme with it's nose to the ground. :sad:

It's better to model a quadruped with it's neck and head straight forward more like this:
http://www.cactus3d.com/HorseBindPose.jpg

Adios,
Cactus Dan

[benytone]

Hi, has anybody modeled and animated a big cat / dog running before? If so, maybe you could help me.
I'm animating a jaguar, I've got a good IK rig and a decent low poly mesh attached using the BIND command and auto weight in c4D v11. However, problems are occurring in the extreme positions of a run cycle...
Mark
Hi, three different quadruped rig and walk;
.
Download: HorseRunCycle (http://e1.video.blip.tv/1010006209230/Magicfx-HorseRunCycleCinema4d967.avi)
.
Download: Jaguar (http://a13.video2.blip.tv/1350000177898/Magicfx-cinema4dJaguar641.mov)
.
Download: Ritter (http://e8.video.blip.tv/1190002704921/Magicfx-RedCrownEquestrianWwwanim3ddetl587.mp4)
.
http://home.arcor.de/anim1d/Horse%20Run%20Cycle.jpg

[Horganovski]

Many thanks for the info Dan, great tip about the neck bind pose.

Very nice work there Benytone, for some reason the Ritter one seems more like a dog than a horse to me, nice cartoony style to it though.

Cheers,
Brian

[jerm]

Just wanted to note, I'm no rigger but I'm good with anatomy, and I'd say the closer you want to get to a realistic cat, the closer your model has to be. You can build a great rig but a lousy model won't be able to make use of it. I'd personally take the model further and the first thing I'd do is look at a hairless cat. It looks to me like you've modeled the outline of a cat with the fur on. For instance, his elbows go right into his body. I know it looks like that when you see a picture, but a cat actually has arms that go up to his shoulders. There should be an armpit there. Their skin is loose as the arm goes to the body, to allow for those crazy stretches they do when running. That's what your model has to be capable of as well.

[Jannis]

http://features.cgsociety.org/story_custom.php?story_id=4020&page=2

Have a look at the link above. The donkey mesh is shown in wireframe with some very close placed loops around the muscle plates the slide and rotate against one another. If you insert these kind of loops and weight them carefully you can have a more responsive mesh in extreme poses and for stretch and squash.

[Horganovski]

Nice link, always cool to see wireframes from movie characters to see how they are constructed.

Cheers,
Brian

[bobzilla]

benytone...with the horse animation, did you do a cycle on the spot then move it through space or did you animate one step at a time.

If you cycled then moved it, how did you go about doing that?

Thanks...

[shibunkin]

thanks all, the response to the post has been great. My conclusion to all the feedback is: think about those edge loops around the hips and shoulders very, very carefully, particularly, as Jerm suggested, separating the limbs more from the torso. Once the edge loops are working reasonably well I'm going to add muscle deformers to the limbs to push out the mesh when the limbs are bunching up.

Cactus Dan, I see the rationale behind the head pose, should the foot/palm pose be treated in the same way? My rest position for the big cat has the foot and hand angled and I was also getting unwanted deformation when those joints straightened out - baggy trousers effect. I was considering recreating the mesh with those joints all lined up - as if the cat was floating and the limbs hanging loose - not load bearing.

back to the drawing board...

[Cactus Dan]

Howdy
... I see the rationale behind the head pose, should the foot/palm pose be treated in the same way? My rest position for the big cat has the foot and hand angled and I was also getting unwanted deformation when those joints straightened out - baggy trousers effect. I was considering recreating the mesh with those joints all lined up - as if the cat was floating and the limbs hanging loose - not load bearing.
The thing to think about is what the range of motion for a particular body part is going to be, and then model that part in a pose that puts it in the middle of that range of motion. ;)

Adios,
Cactus Dan

[benytone]

...for some reason the Ritter one seems more like a dog than a horse to me...
Brian
hehe...

benytone...with the horse animation, did you do a cycle on the spot then move it through space or did you animate one step at a time.
If you cycled then moved it, how did you go about doing that?
Thanks...
hi, thats just Converting keyframes into Clips in the NLA Editor
Small sample: Download preview (http://a11.video.blip.tv/1790006695166/Magicfx-Cinema4dNLASample462.avi)

.

[benytone]

sorry wrong...!

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.