Manuel Bastioni Lab, turns Blender in a laboratory for 3d humanoids creation

I am using Blender compiled as a module and importing the lab.

Whenever I do proxy fitting I get lots of errors regarding “Error: Vertex not in group”, but I do not understand why or if I can somehow filter it out.

Does anyone know how to fix this output or ignore it?

When I try to change expression for Caucasian female to “sleep”, only one eyelid gets closed. Not sure if its a bug or if I am missing something.

https://imgur.com/IJYW58n

Some good videos for importing into mlab characters into unity on that site.
I will keep it in mind once I familiarise myself with UE4 a little more.

Pavs, NumesSanguis, Thank you!

Sorry for late reply.
The most reliable way to turn the model in a static posed character is to apply all the modifiers.
Remember to always save a copy of the poseable character version before proceeding.

What version of Blender are you using?
What do you mean with “Blender compiled as a module” ?
What version of lab?

I am using Blender 2.79, compiled into a Python module as described in this post.
Using lab version 1.6.1.

Thank you for a great plugin and quick responses!

This might be slightly off topic, but Manuel have you seen this?!

They talk a lot about materials, rigging, facial rigging, touch base on Mocap, wrinkle maps.




1 Like

[ATTACH=CONFIG]519239[/ATTACH][ATTACH=CONFIG]519240[/ATTACH][ATTACH=CONFIG]519241[/ATTACH]

[ATTACH=CONFIG]519239[/ATTACH][ATTACH=CONFIG]519240[/ATTACH][ATTACH=CONFIG]519241[/ATTACH]

I cannot find the string “Vertex not in group” in the lab messages. Is this the exact error message that you receive?
If yes, it’s not from lab, but from the Blender system, and I cannot reproduce it.

Anyway the lab is not designed to work in “module” mode: it requires interaction with the user and other things. Trying to use it as module will probably return many errors.

Thank you. It’s a very interesting challenge to design a system that can optimally work on Blender and also on external platforms with so many differences in performance and features.

In case someone missed it, here is the article with more details about the design of lab 1.7.0 and the intermediate releases: http://www.manuelbastioni.com/wip_170003_new_models_gui_and_expressions.php

Hi Manuel

The new teeth model looks very good. It should be detailed enough to make actual dental prothetics from. :cool: It’s great to have this much detail in the mesh itself without relying on displacement maps.
http://www.manuelbastioni.com/gallery_170_03.php

I have once again a special request question. I was thinking about the implications of the new topologies now becomming more tailored to specific character types, and I was wondering about how easy or difficult it will be to create combinations.

Say I want to create a realistic looking character but with uncommon body composition. E.g a character with very skinny weak legs but with a strong athletic upper body (wheelchair bound), or a character with emaciated arms and legs but a very bloated stomach (sub-Saharan severe malnutrition).

Will there be some core topology edges and/or poles that are shared between all the realistic human topologies? For example, could I create one skinny character and one fat character, delete the mid torso of the skinny character and take the mid torso from the fat character and merge it into the skinny character (skinny character but now with the actual fat stomach topology) without introducing any new poles or other undesired complexity into the combined topology? Obviously the merged area would need to be smoothed and tweaked a bit afterwards, but would the topoogy itself still be clean?

I am not really concerned with the practical complications and technical difficulties at this time (armatures, shapekeys and whatnot), only the topologies themselves, and to what degree the major body parts (head, chest, stomach, pelvis, arms, legs, hands, feet) of the different topologies might be interchangeable. Again on a PURELY topological level. For a possible future application where this ability could prove useful.

Bodypart-separating lines might not be clean simple cuts, but follow the natural flow of the body topology. The lower edge of the chest on this page illustrates what I mean.
http://www.manuelbastioni.com/gallery_170_01.php

In the other hand, non-complex bodypart-separating lines, simple edge loops without any poles (say around the upper part of the biceps or thigh), might perhaps be easier to maintain. I don’t know.

Of course there are always ways around this mixing regardless of the topologies being merged. If all else fails resolve to sculpting and manual retopology afterwards. But I think it would be very neat if the new topologies allowed for this type of interchange.

This is not something I expect will be possible, but seeing the thoroughness of your approach, I nevertheless present the idea for consideration, so you can evaluate whether this special scenario is something the new topologies could or should support, or if it introduces too many limitations on the specialized topologies.

Again thank you for your work and dedication to this amazing project. Really looking forward to 1.6.2 and beyond.
http://www.manuelbastioni.com/progress162.php

if I am not mistaken, this release will also have simplified meshes for game content creators?

i use 1.6.0 meshes for my game and they look and perform awesome. just dont subdivide and delete all the cloth covered verts.

they way i see it, any lower resolution, and you start to lose the awesomeness of using mblab meshes. normal baking can only do so much.

my characters average 15,000 verts (accounting edge splits) with clothes/armor.

Is there a way to tilt the character’s head with IK? I can’t seem to find a control for that. The only thing I can do is control the LookAt thingy.

First at all sorry for my english.
Quality of models much better than makehuman. Makehuman female models with large breasts or muscular character are square, pronounced low-poly.
Polycount bastionilab acceptable for modern game engine’s, but news about new teeth with 11000 faces really frighten.
I hope that the promised low-poly models will not just be with fewer polygons but the same quality / polygons as now. Quality makehuman is not enough for modern games. This is LOD, not main character quality.

When animate bastionilab character i have some freezes, not smooth. IK configured incorrectly, often the character’s arms and legs are turned out at unnatural angles. Maybe freezes my problem, and IK glitches because freeze. Makehuman with IK-rig work normally, smoth and correct angle\pose.

Wishes:
Make face and head settings in one panel. More face settings, even a random generator does not change the face too much now. Manually I did not really succeed too. Save\load face and head settings separate from save\load body settings.

Wishes for far future:
With the help of neural networks and AI some programs are able to generate a three-dimensional model of a face by one photo.
So we can get an approximate face shape from one photo.
The neural network could be played with user-accessible face adjustment sliders until it gets about the same shape and overlays the texture with the photo. It would be just awesome, killer-feature.

The same could be done with the figure of the body, because it is not so difficult to get from the photo \ sketch silhouette, which we want to repeat. And approximate with sliders.
Yes, I know that about the body figure there is a lesson with automodelling tools. It’s close, but not quite what i want.

Given your knowledge of human proportions, I think it would be possible to determine the approximate shape of the body from one photo.
If the belly protrudes into the profile, it is logical to assume that it will be rounded in front projection.
To build such assumptions work just for neural networks.

This is for far future, when you suddenly think that you have already made an ideal and complete character generator.
To make it so that this generator could create a model by one picture or picture would be just wonderful and would solve many problems of lazy indie game developers.
The next step would be to remove the skeleton animation from photos or videos.
Like this.
http://https://www.youtube.com/watch?v=eUnZ2rjxGaE

A small teaser from my project, FACSvatar, for animating the facial expressions of Manuel Bastioni characters:

The data for the facial animation can come from any video with a person in it. In the above video, the facial expressions are extracted as FACS data, then converted to Shape Keys, and finally send to Blender, where they are saved as Keyframes. So afterwards editing / exporting is possible as well.

All of the code is open source and can be found here: https://github.com/NumesSanguis/FACSvatar/ (at the moment in the Blender branch).
Sorry, the documentation is still a mess, but this should improve soon. So I recommend waiting a bit before you plan to use it.
The homepage: https://surafusoft.eu/facsvatar/

You can also directly stream the data into a Unity 3D project with a MB character in it with the FACSvatar framework.

Hi vida_vida,
the creation of common edgeloops between body parts is not easy. It enforces to have the same number of verts along the flow, and this can thwart the optimization of the mesh. There are also “anatomical” problems, for example the direction of obliques muscles is very different from the ribs direction, so defining a common chest line between a bodybuilder and an underweight model is very difficult.
Also keeping a constant number of verts for so many loops (neck, breast, shoulder, pelvis ecc…) for all base meshes will create too many constraints in the modelling. I will try an experiment, but probably it will be not feasible.

I’d like to add low poly models in 1.7.0, but I have not yet officially added them in the roadmap because the amount of work for 1.7.0 is already enormous and perhaps low poly will be postponed to 1.8 or to one of 1.7.x releases. “Stylized” is referring to anime style, not the topology.

Do you mean side tilt? No, at the moment.

Amazing!

“Error: Vertex not in group”

I have the same errors when using blender in background mode. The addon works but the proxy fitting function prints tons of errors in console. You have to disable standard output before calling that function to make it work.