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

Wow, this addon looks really amazing, great work.

I will definitely have a look at it :yes:

Those are a basic set of shape morphs. Makeuman has a comparatively elaborate set of shape morphs for certain body parts, particularly the face and head. Iā€™m not requesting anything though, as I understand that you are alone with this project. Youā€™re doing it for free, and because of that it is in your spare time. I was just highlighting that weā€™re pretty limited in shaping our characters (except the manga ones which obviously need no such granularity Iā€™d assume). We probably can resort to sculpting for now I think.

Can you tell me a practical example?
MakeHuman has an advanced algorithm to mix the macrodetails only (gender, tone, weight etcā€¦) but the face details are just simple morphings.

ManuelLab include some advanced algorithms not yet ported to MakeHuman. For example the algorithm used to calculate the breast tone/mass is very different from the one used in MakeHuman.
Also ManuelLab is more accurate in face morphing, because it uses different set of morphings designed for the specific characters.
For example, in ManuelLab the asian eyes have a dedicated set of morphings since they are very different from caucasian eyes.
In MakeHuman the same morphing is used for both asian and caucasian eyes, and this can create some problems for extreme values.

Wooops! I hadnā€™t used ML since last time I commented here (not today, another day), and only remembered the face morphs with only three sliders, and another face section with not many sliders (I should have explored more). But as I check again ML really does have an extensive set of shaping morphs for the face afterall. I simply assumed the eyes and nose forexample hadnā€™t much morph sliders either. Assumption must be the last resort, sorry about that :o

The 3d model of male anime for next version of ManuelLab is almost ready.
More info: http://www.manuelbastioni.com/wip_110002_anime_male_base_character.php


The next version of Manuel Bastioni Lab will improve and complete the support for anime characters.
I will include the male version for the ā€œshojo classicā€ and I will add a completely new style: the ā€œmodern shojoā€.
The image shows the main differences between the classic and the modern template.
Higher resolution available here: http://www.manuelbastioni.com/gallery_wip_110_003.php

Also I just started an official facebook page, in case you want to follow the development:
https://www.facebook.com/manuelbastionilab


Just been plaing around with thisā€¦ simply amazing!

Thanks for the such an amazing addon! Not only is the addon great, but the presentation is also very appealing. And the planned features are exciting!

3 things for me:

  1. One thing that saddens me is the lack of a more intuitive rig. But, while I didnā€™t see anything in the planned features page, I did see you reply that ā€œanimation controlllersā€ are planned. Would that include IK controls and a facial rig?

  2. I love the pose/expression library, and while the selection is good, I wonder if itā€™d be possible to have the ability to add/remove poses and create custom libraries?

  3. Whatā€™s the ā€œsmart auto-modelling systemā€? Is it a system to easily customize various aspects of the model with perhaps sliders? Also, would this enable creation of, say, teenager or children body types?

I apologize for the bunch of questions, haha. Just very interested in the planned scope/direction of the addon.

Grazie mille Manuel! Itā€™s amazing piece of code and art. Models themselves are just awesome! And as my dark (geeky) side is telling me, the script handles boundary boxes for proxies like a charm. Iā€™ve found it much better to fit garments this way instead of parenting them to the armature with weights.
Iā€™ve missed some functionality (just for control flexibility), and Iā€™ve added it to the script. Now it does handle multiple humanoid instances in the scene, can switch between them on the fly, and allows to create new ones on demand. Also Iā€™ve added option to limit meshes affected by Store proxies / Fit proxies operations with selected ones only. Itā€™s very handy if you try to fit a lot of items to some different chars changing their poses.
Changes I made are extremely lame because Iā€™m new in blender and in python as well :slight_smile: But it works for me. I think itā€™s better to ask author before sharing my hacks. So, Manuel, can I share it here or somewhere else, or send it to you to look over it first?
Here how it looks in GUI:



Hi 0o00o0oo,

I missed to add it in the public todo. I just fixed the page (http://www.manuelbastioni.com/guide_features_planned_in_next_versions.php). Facial rigging is not planned atm, since I prefer to develop more the shapekey solution, but IK and controllers are already on my personal todo list.

You can already add custom expressions. Just create your expression on a base character and export it as json. Then put the generated json files into the manuelLab addon folder. For example if you create an expression for caucasian female, you have to put it in your [ā€¦]scripts/addons/manuellab/data/humanoid_humanf01/expressions folder.
In the same way you can also expand the character preset and the phenotypes.
Custom poses are not supported yet.

The automodelling is a system that start from a rough input and return a realistic character. I wrote a prototype 7 years ago, but it was never ported to MakeHuman: https://youtu.be/J1-5Y4RAZR4

You are welcome! I created a facebook page and a twitter page to post the daily progress. The Facebook one is https://www.facebook.com/manuelbastionilab

Hi Mechanic, thx for the experiment.

I intentionally carefully designed the software to work on one character only. There are some reasons for this, basically all them are related to have an hyper-stable, error-proof program.
Having multiple objects with different topology in the same scene can create problems with the Blender undo system, can confuse the script and potentially can crash it.
Also I will add more complex algorithms in future (automodelling, skin generator, ecc), where it will be very important to have only one manuelLab character in the scene.
For this reason my script only consider one character, even if the scene contain more humans.

Best,

Manuel

Manuel, thanks for explanation. Itā€™s quite good intention to keep software bug-free and reliable, but assuming that any new feature can cause a bug we can avoid any new code at all! :wink:
I just think that if we properly isolate the current the_humanoid instance for the script to work with, and just follow some simple convention for resources usage within the script (actually filenames) for different instances in the scene, we have too little chances to confuse the script. Thereā€™s no need to handle any kind of array of Humanoid instances simultaneously, the question is just to allow the user to re-init the existing single instance with another blender object. All the rest of the functionality just uses the Humanoid.get_object() and canā€™t use something unrelated by definition.
Perhaps the current approach to polish the character outside of the real scene context is great for academic scientific needs, but itā€™s hard to use for artistic scenes where must be more than one character with some kind of relations between them. :frowning: For example how one can handle a simple scene with a hugging couple using current single-instance workflow? Of course itā€™s possible to keep 3 different files (one per char, and the scene), and then sync them together after any change on char, but it seems that having all the scriptā€™s power for any character directly in the scene is more natural and easy.
Maybe the real pros are more significant than potential cons with a little effort. Anyway, you are the one who is an author here and Iā€™m just a user with my feature request, so itā€™s all up to you. :slight_smile:
Iā€™ll try to show my little hack just as a proof of concept. And it would be great if such a change will not break the code structure and would be easy to implement and stable to work.
Thanks again.

Yes, I see it now, itā€™s bad idea to use it in the working scene, considering future road-map. Itā€™s really better to have less bugs in the core than more top-level features. Sorry for my molestation.
And I have a little idea for future releases (again :-)). Now each new humanoid has normals auto-smooth for 3 degrees enabled by default. It causes a lot of flat faces on render with surface subdivision. It can drive nuts someone who never used auto-smooth before (like me :-)). So, maybe itā€™s better to turn it off by default to keep some not-experienced usersā€™ mind intact?

Hi! And by the way congrats to Manuel Bastioni, his work really does look awesome.

Iā€™ve had the very same problem when trying to create my own rig of the model, and after some investigation the problem seems to come from the eyebrows and eyelashs: if you manually delete them, you can parent the rig with automatic weight just fine.

Now, if someone more experienced than me had a nice solution to this that allows to keep the eyebrows and eveything, they would be very welcome!

Thx, I will disable the autosmooth in next release.
Anyway I only know the autosmooth in Property panel -> data tab. It has only one parameter: the angle. Where are the 3 degrees options?

Instead of delete them, you can just detach them using the Pkey in edit mode. After the autorigging, you can re-join them with CTRL-Jkey, and paint their weights by hand.

Thank you! I canā€™t wait to try this!


ManuelBastioniLAB 1.1.0 ready for download!
http://www.manuelbastioni.com/release_note_110.php
Enjoy!

Thanks for the update.

What were your references for the anime models? To me it looks like the legs are really long and the head is a bit small. At least for a modern style I would prefer something like this:


(taken from pixiv)

There are infinite variations in the tree of anime/manga styles.
The important is to identify the basis for the main branches (this is what Iā€™m doing).
For example the images you posted above can be easily obtained as variation of the classic shojo style, just modifying the parametrs for legs and head).

When a style is too hard to be reached using the current base models, this mean that we need a new branch for it.
For example, in next versions I will add a new base for game-style, west-style, and so on.

With Legs_LowerlegLength and Legs_UpperlegLength at 0 using the classic shojo model it was pretty close, but I still think the default legs are a bit too long. At least I havenā€™t seen any anime characters with legs that long. :stuck_out_tongue:

Some other issues I found, the addon doesnā€™t respect scene unit scaling so if you use something other than the default (like 0.01 metric scale for correct export to the unreal engine) the character will be really tiny, 1.8cm or something. Even if you resize it and apply transforms the moment you touch any parameter strange things will happen. I guess you can finalize the character and then resize though.

Also I donā€™t see how you can apply the subsurf modifier while still being able to generate shape keys for the expressions, also for exporting to a game engine. If you apply the subsurf before finalizing you canā€™t actually finalize, all the options disappear. If you finalize without applying the modifier you canā€™t apply the modifier at all because of the shape keys. Maybe this addon isnā€™t aimed at game developers though.

If it is, something that could be useful would be to add simplified shape keys for some of the parameters so you could use in character creation. It seemed like parameters had multiple shapes so maybe this is difficult.

Anyway, a very interesting addon.

For real people, having the legs longer than the trunk is a beauty factor. For anime this proportion is exaggerated. On the contrary the ratio legs=trunk, as the one you posted in your reference image, is rarely used in shojo, and more present in kodomo.

Some other issues I found, the addon doesnā€™t respect scene unit scaling so if you use something other than the default (like 0.01 metric scale for correct export to the unreal engine) the character will be really tiny, 1.8cm or something. Even if you resize it and apply transforms the moment you touch any parameter strange things will happen. I guess you can finalize the character and then resize though.

I will investigate about it.
Anyway the addon is designed for a different pipeline. Itā€™s not to work directly in the final scene, but to create the character, save it in a separate file (creating a character library) and then import it (after finalizing) in the final composition scene.

Also I donā€™t see how you can apply the subsurf modifier while still being able to generate shape keys for the expressions, also for exporting to a game engine. If you apply the subsurf before finalizing you canā€™t actually finalize, all the options disappear. If you finalize without applying the modifier you canā€™t apply the modifier at all because of the shape keys. Maybe this addon isnā€™t aimed at game developers though.

Most of game developers are happy with 15000 - 20000 vertices for a single character.
Iā€™m receiving many requestes for low poly versions of the base meshes.

If you really need, for your game, a character with about 70000 vertices, you can create different versions of the same character, one for each shapekey (applied), then subvide all them, and then use the ā€œTransfer shape keyā€ standard Blender function to ā€œmergeā€ all the shapekeys to one character.

In case you need the subdivision to sculpt the mesh and create detailed normal maps, the shapekeys are not needed.