WorlForge Standardized Humanoid Animations Rig and Library

[Demarii]

I've posted a video which discusses the new animation rig and library I've been working on for the WorldForge project on youtube. Since the Ember game Client for WorldForge uses Ogre3D engine and in keeping with the spirit that is Worldforge, it simply makes sense to share this information with the Ogre 3D community as it could prove useful to other development teams who need to create animated humanoid assets for their own projects.

You can find it here:

http://www.youtube.com/watch?v=i-jm4Lsbwkk

[youtube]i-jm4Lsbwkk[/youtube]

Now that I have settled on the Standardized Rig format, and have a decent workflow in place I expect to make much greater strides in the creation of the Animation Library. My plan is to create a series of wiki pages, and video tutorials which will guide world builders, and 3D artists in the different stages of creating and working with the animation data so that they can make the best use of the material provided and should they choose to help by extending and improving the data sets as well.

Once the base set of data and the workflow is well ironed out. With the Wiki pages and videos made completed. I would like to pursue possible scripts and adaptations which could better allow for refining and working with the motion capture data, and exporting the data into a complete game ready format in the most effective manner for 3D artists. Since I'm learning a lot as I go I'm always happy to receive any comments, suggestions, which will help me in improving these assets for everyone.

The data files will be made available on the media repository once I have completed work on the base default animations and have fully working assets that replace the existing game content completely.

I will periodically update this thread as new wiki pages, videos or data is made available.
Enjoy!

[ppClarity]

License? GPL/LGPL is a non-starter for a lot of people.

[duststorm]

Interesting project.
I'm also experimenting with creating an animation library, but centered around MakeHuman.
You really have a nice mocap setup (6 camera's) What type of cameras are they? The place where they are set-up, is it a university?
I'm limited to one kinect and the freely available motion capture data on the net.

Lots of succes with your project! I have always appreciated how the worldforge project made its resources available freely.

[Demarii]

@ppClarity,

Yes all WF media is GPL licensed. I realize this won't be to everyone's liking, but for some projects it could be of use. The goal of the system is to provide a complete easy to use asset set for our project and anyone can make use of it is more then welcome to take advantage of it however they choose.

Now that being said, the WorldForge Rig and animations library built has a jaw and hand bones which will be articulated, cleaned of errors and then eventually improved over time for use by all our humanoid characters. But since I am the creator of the original mocap data, one thing that I can do is to release the original raw mocap data in its standard BVH structure before it is transferred into the WF rig format under a different license then GPL which would allow you and any others to make use of the same base mocap data that the WF rig uses and build from there in the same manner. I'm currently talking with the folks over at http://opengameart.org about letting me submit the mocap data there.

@duststorm,

Thanks! Best of luck to you as well. I really like MakeHuman and would love to make a Proxy and associate the rig with it, but that will be something to look at in the future as mocap and a few stages of media upgrades currently take up the majority of my free time. Actually my studio is the largest room in the basement, the only thing I would change is the lighting, the pot lights and lower ceiling tend to make the upward facing cameras light filter and they lose a bit of the color detail as a result hurting tracking a little. I use 6 PlayStation 3 Eye cameras for larger captures of complex actions or tasks. I really like and tend to prefer the dual kinect camera setup, two is really very nice. I looked at the mocap available online but it really did not suit our needs, so I decided to start producing my own that would be more focused for the Mason game system we are developing. It's a game of pioneering and as such things like chopping trees, working wood, digging in various forms, farming tasks, and general building actions are at the top of the list along with general movement and gestures. Eventually I would like to continue to expand to include more variants of the main animations to better represent different body types, and also more special cases like combat actions etc.

[duststorm]

I assume you're using the Ipi soft software for motion capturing?
Unfortunately I only have one kinect and have experimented with a few projects like Brekel and ni-mate, but it's really not good enough for capturing anything serious. Even very basic animations are very jittery. Oh well, with some cleaning up they can still give you a headstart over manual animation (oh I hate that so much, it takes me such a long time to do a convinving walk or run loop)
Does it really work well with 2 kinects? I might even consider buying a second one someday

[Demarii]

Yep I'm using IPi Soft tools. I really did not like or use Kinect until I had the ability to use 2. I just didn't find the resolution that good you always had to be facing the camera. With to I can do quite a bit with it, and prefer it over the 6 cameras when I can use it because its so much faster of a workflow.

A good portion of my time no matter if I'm using 6 Camera or 2 Kinect is after my initial processing going out setting my "Takes" and then post processing those to fix errors, and clean up any jitter or trajectory issues. I prefer to do that on a take by take basis because different actions will need different smoothing factors applied. The MoCap I create is generally very decent even in its raw form it can be broguht into the client after some basic orientation and translation fixing. Normally once I get it to blender I want to make some improvements like bring the elbows away from the chest a bit in a walk, or fixing a rotation error in an ankle, but overall most of it is pretty clean like you say to start from. Really I find it to be cleanest and smallest format to build a library of multi-purpose animations from. As you work on a small piece and get it looking good you export it and you have a new more updated version as your base, or possible a variant to work forked.

I'm like you not a huge fan of doing all the manual animating I just don't think what I come up with that great, I find its so much nicer to work from a realistic base even if it has errors its simpler to fix those and add little touches then build from scratch. Not to mention you can keep your bone counts alot lower as you don't need a lot of IK drivers.

2 Kinects is honestly the only way to go, In the video I show how you can create a 360 depth capture of yourself, basically a hologram to capture actions. Its pretty sweet and can handle lots of static types of animations, I've even been able to capture decent swimming action using a weight lifting bench to lie on.

[jacmoe]

Guys, we've had this discussion before:

Art licensed under the GPL is not viral the same way that code is viral.

So, according to the FSF, you can safely use GPL'ed art in your commercial project.

I asked Azathoth (project lead for Ember) and he asked the FSF.

So, yes: this is really cool.

Please update this topic, as I've often wanted to make use of WFs humanoid models but never had the time to figure out how.

[Demarii]

@jacmoe, that's really good to hear! I've not had a chance to chat with Azathoth recently as its a busy time of year with GSOC starting up etc. But I'm glad to hear he looked into this as I was under the impression it was not as bad as some believed it to be either but not being a lawyer or knowing for certain I thought it best not to get into it. Ogre users will then be able to make complete use of the this library, rig and workflow.

I will most definitely update this thread when the library is brought online with links to the data, or I've created any new tutorials or videos for people to benefit from.
Thanks for clearing that up.

[ppClarity]

When I read "library" I read it in the context of it being code. Is this just a library of art assets or is there a functional portion?

[jacmoe]

It would be great if the library - if it means code to handle the rigs - could be licensed under a GPL-compatible non-copy-left license, like MIT.
But I understand if you can't or won't do that.
Some of us can then come up with our own WF-Rig handling code in that case.

The rigs themselves are GPL.
Which is - much to my surprise - perfectly cool for any kind of project.
You need to point people to the source of the models, and contribute any changes back, or make it available to the public.
The regular GPL-business..

[Demarii]

@ppClarity, yes this just concerns a collection of assets. There will be number of files for artists to use at different stages in the process. They can either build their own skeletons from the bvh's or they can simply bind their meshes to the standard rig and use the standard skeleton provided.

I use the term library in its traditional sense.

@jacmoe, Azothoth is handling the code part of things and it will likely be part of Ember which is gpl. Currently it is defined in xml at the model definition level but i think once we have settled on the parameters and blending methods in client editing functionality will be added to the model editor widget. Other then that it stock ogre animation functionality as far as i can tell.

At some point we'll have to work out something for pose animations for humanoids as well, but for now focus is on a very functional and streamlined skeletal based system.