Blenrig?

[Somguy]

After seeing this, I immediately thought makehuman.
http://www.jpbouza.com.ar/ESP2/descarga ... ig-3/id/en

It would seem that makehuman, and blenrig have the similar goals. Blenrig uses muscle deformation for realistic posing and animation along with an extremely detailed and useful rig. Because the rig is apparently mesh independent.
http://vimeo.com/5092489

Considering the similar goals, and the opposing strengths (Makehumans customizability versus Blenrigs realistic muscle deformation system) it would seem like a collaborating made in hi-res posing heaven. Now, obvious the issue is the modifiers, and exporting them in a useful form. I suppose it comes down to the modifiers (since blenrig is hugely dependent on them).

Thoughts? Is there some blaring fault in my suggestion?

[Manuel]

Thanks you !!!
I'll talk about this with Thomas.

Best,


Manuel

[cousteau]

Another interesting video:
http://www.blendernation.com/2011/05/21 ... or-bipeds/

Maybe MakeHuman could use BlenRig as a back-end for the posing system.

[ThomasL]

There are certainly some interesting techniques there. The Blenrig is certainly worth studying. I did have a look at version 3, but the new version seems a lot better, and especially faster.

One serious obstacle is that Blenrig uses the mesh-deform modifier and hence requires a mesh cage. It is possible to configure the mhx exporter to export a cage (http://sites.google.com/site/makehumandocs/blender-export-and-mhx/mhx-import-in-blender-2-5/cage-and), but unfortunately the cage is of poor quality and need a lot of fixing before it can be used. The clothes-creating algorithm (http://sites.google.com/site/makehumandocs/blender-export-and-mhx/making-clothes) works quite well for proxy meshes and snugly fitting clothes, but rather poorly for a cage which is supposed to float at a distance above the skin.

If you want to use Blenrig and MH together, you can try to clean up the cage that MH exports, but perhaps it is faster to skip the mhx rig entirely and retarget Blenrig as is described in one of JBouza's tutorials. Either way quite a bit of manual labor is necessary. To get something that works out of the box, and works well, one probably needs to model a low-poly cage mesh for each MH target, but that is probably not an option.

[jpbouza]

Yes, manual work would always be needed. Specially if the character changes its shape too much... Changes in the length of the body parts wouldn't be such a problem, but if the character gets fatter or thinner, then you should probably be forced to tweak the Mdef Cage...

The only things I can imagine for accelerating the process would be:

1) That Make Humans characters had already their weights painted and ready for the bone namings of the rig (as shown in BlenRig 4 tutorial 3)

2) That there would be an automatic way of retargeting the rig... according to the character's proportions in make human...

3) if this automatic way existed, the Mdef cage should be somehow generated with the new proportions of the rig...

I'm not a coder, so I can't do all these things, hehe.
I don't know, for now, with the method I show in the videos, it just takes a couple of hours (or less) to rig a character... Once the character is rigged it is really easy to reshape it and create new ones...
It would be great to have more integration with Make Human or other Blender's rigging tools, but it would require coding I guess...

[jcapco]

We are writing now our own pose system that should work similar to mesh deform modifier of blender. Except that the user won't be able to modify the weights at will, it will all be computed from the joint cages. Our cages are a bit different, the links (part that is between two joint or between a joint and an end effector) don't have cages but the joints do have it. Each vertices in the link is thought to have a weight of 1 with respect to the rotating joint and no other control point. I did this because the problem that I saw with volume preservation upon joint rotation was seen in the joint area of the body. The link almost never deform in shape (except rotation and translation), it only does that upon self-collision which is another problem in itself.

So once our pose code is finished you already have automatic weights in the human vertices according to the right human target. This could be exported to blender, but then you only get a weight object and no cages.

[Sethren]

Do you guys actually have self-collision solved or close to being solved in the new pose tool?

[Manuel]

Do you guys actually have self-collision solved or close to being solved in the new pose tool?

Not exactly, but there is an interesting workaround for static poses (not custom poses).