Trouble with Retarget BVH

[tomcat]

But wait? You work with a translator? For everything? I didn't notice that, at all.

Yep, I have a very poor ability with languages. One of the main reasons I haven't evacuated yet. I recommend it — DeepL

You need to repeat this for the other shoulder bone and the upper arm bones

Now I kind of get it. If the shoulder tilt is corrected, then the arm will follow.

Yeah, a few animations can be changed that way, but if you have to do it en masse… it gets a little sad.

But why did you say

bone roll of the legs

specifically the legs?

[PayItForward]

But wait? You work with a translator? For everything? I didn't notice that, at all.

Yep, I have a very poor ability with languages. One of the main reasons I haven't evacuated yet. I recommend it — DeepL

you already told me about DeepL but I had no idea how good it is. It's actually a bit scary.

If the shoulder tilt is corrected, then the arm will follow.

Yes, you see it in the picture. Correcting the shoulder will force you to correct the upper arm as well. And then there's still the problem with the neck. I personally hope somebody will find a technical solution, though.

a few animations can be changed that way, but if you have to do it en masse… it gets a little sad.

That's the problem, yes. For me it's bearable since I don't need that many animations but CMU has like 3,000 animations. I have downloaded the videos http://mocap.cs.cmu.edu/allavi.zip and watch what I need before. So yeah, it's all about making good decisions beforehand.

legs

Well, not specifically. But I have encountered the problem mostly with legs that rotate completely unnatural. Maybe other bones are also wrong but perhaps just not noticable.

[PayItForward]

I got a bit farther with Retargeter BVH. After some further testing I'm even more convinced that the root of the problem is the T-pose that was added later to the CMU animation data. More specifically, it isn't even a true T-pose but a mixture of a T- and an A-pose (see image: C).
That's fine by itself, btw, but the problem is, the more this pose differs from the 'rest pose' that comes with Makehuman (i.e. usually image: A) the more incorrect the result will be.

So to get a perfect result you basically need a 'rest pose' which is virtually identical with the 'rest pose' in the CMU animation data (B=C!). Since Makehuman right now has no pose that comes close to the pose needed I made one on my own (see image: B).
It leaves still a lot to be desired. There are issues with the neck in some animations, also with the toes, and the shoulders. The shoulders sit so deep in the CMU rig that I don't really know how to reproduce this effect in my pose.

I attached my T-A-pose so you can test it if it works for you.

Now I'm going to test if the same is true for the ACCAD male BVH.

[tomcat]

I attached my T-A-pose so you can test it if it works for you.

Well… it works!

Blender 3.4.1

It's actually a bit scary.

In fact, I'm editing the translation a bit. Although I have no aptitude for languages, I have studied Polish and Hebrew… and even the language of the island aborigines… which the world is trying to speak.

So yeah, it's all about making good decisions beforehand.

That's probably the hardest part… one never knows in advance which decision will turn out to be the good one.
__

By the way, an interesting plugin for Blender, in case you need it.

[tomcat]

Some reflections on ancient problems. That the poses and movements of the characters would be more natural. The idea was this.

1. Rig should take into account the size of the joint (bone thickness). A similar thing is done in Z-Anatomy. BIOMECHANICS.

Here is the MakeHuman rig:

And this is the rig of an anatomically correct skeleton:

My suggestion is to add bones that would compensate for the size of the joints:

• Rig MakeHuman
  
  

  

• Rig with joint compensation
  
  

  

For a correct rig, you need to consider the size of all the joints. For a light (gaming) one, I think the elbow, hip and knee joints would be enough.

2. The presence of "control bones". Similar to those in Biomechanics blender:



Bending the knee or raising the arm causes movement of connected bones that move together. For example, raising the arm causes simultaneous movement of the shoulder and collarbone:



3. And, of course, the spine, which consists of several vertebrae.

4. And a slight movement of the jaw sideways.

Unfortunately, this is really a lot of work… perhaps even too huge.

It is impossible to do everything at once and perfectly… but you can try to determine the vector of development and lay the foundation.

[RobBaer]

Really interesting insights, and well illustrated. Thanks for sharing.

[tomcat]

@RobBaer Thanks for the response! If my points were taken without negativity, I'll continue my thoughts.

Soft tissue muscle and fat — rig or shape?


I tried to find the most decent picture on this subject.

Muscle work and fat sway in motion. No revelations, just a comparison of known solutions.

Rig (extra bones):
+ easy to implement, require few resources when used in animation (which is important in games).
- limited in possibilities.

Shape Keys:
+ more morphing possibilities.
- difficult to create (sculpting targets).

Different situations benefit from different approaches. In my opinion, one of the main advantages of MakeHuman is variability, adaptability to different, often conflicting requirements. It is possible to choose the detailing of models or rig more suitable for a particular purpose. Therefore, it is desirable to leave this approach and let users choose.

[PayItForward]

Although I have no aptitude for languages, I have studied Polish and Hebrew… and even the language of the island aborigines… which the world is trying to speak.

Now that's quite a combination

That's probably the hardest part… one never knows in advance which decision will turn out to be the good one.

Just ask me how many pages I've written that never made it into the final book.

biomechanical skeleton

Hard to imagine seeing this anytime soon. I was always hoping somebody would make something in Blender similar to 3Dmax' Bipeds with physical correct bone restrictions but there really seem to be larger issues making that unlikely.

[tomcat]

biomechanical skeleton

Hard to imagine seeing this anytime soon. I was always hoping somebody would make something in Blender similar to 3Dmax' Bipeds with physical correct bone restrictions but there really seem to be larger issues making that unlikely.

Well, it's already done in Z-Anatomy - BIOMECHANICS (link at bottom of page): proper bones and constraints. The problem is to make children's skeletons, for growing characters.

[PayItForward]

Well, it's already done in Z-Anatomy

oh yes, I noticed. It's great but I meant something like that in Blender. I tried to make a skeleton with natural bone restrictions in Blender years ago but it never turned out good. Would be fun to have it react to the physics engine.