I've been looking through portfolios and it seems like the ways people t-pose their characters varies a lot. Most for the most part it's arms out at a 45 degree angle. But for example in the 3 point portfolio the characters all have a slight elbow bend with the hand kind of curved at the palm with fingers and thumb facing out. Then there's Mashru Mishu's work where everything seems to be an ideal neutral standing pose.
Are there advantages to posing characters like this or are these variations mostly done out of sculpting preference?
Replies
The other poses have come about primarily to help with animation weighting and deformation.
Any deforming body will have some level of stress or compression through the muscles and skin and clothing etc, to hold the pose. The ideal neutral standing relaxes most muscle tension, and so should deform decently well once rigged.
The other prime concern is virtual bone structures often have prefered axis of rotation. This either has to be manually set, or inferred from the initial rotation of the bones. So if a bone chain is straight along one axis, no preferred rotation axis can be inferred, making for more setup time, or worse, strange animation behaviours.
Also of note is that triangular geometry stretches more believably than it compresses.
Modeling with every joint slightly bent in its most common compression direction addresses both of these concerns by allowing the initial pose of the skeleton to detect its preferred directions easily, and fake proper skin bunching when the skeleton deforms the mesh into a compressed state.
Practically speaking there is an argument for a bias towards what will be most usually seen, out of the total range of motion. We could excuse a less good looking arms-right-up pose if it is only ever seen once in a blue moon, and the remaining hold-gun-and-look-awseome animations look that bit better.
One thing about tposes/reference poses is that the surface is also usually defined in the tpose/reference pose. What I mean by this is the vertex normals/tangent basis will set in this pose for deformation further down the pipe. Normal maps should threfore be baked or at least tested to look correct in this pose, as it is the best they will look, every other pose will introduce some level of distortion.
Then again, it is sunday morning so I could be talking hot air.
It's simply unatural for humans to have perfect 90* angle in hands, arms and feet, it simply doesn't make sense.
Plus, with armor type design on a character, it can become a hassle when trying to position the pieces to fall one under another.
So unless you can predict on how to make a 90* character not have their biceps bigger then their body when deforming, or armor pieces that will clip through the character, choosing a more relaxed pose helps alot more in the limitations and design.
Either way, it's all about finding the proper topology through trail and error, and having enough polies in said area so everyone is happy.
So: If no extreme movements are to be applied to the character, 45° will show better deformation results and is easier, or makes more sense for modellers (i'm guessing here). In rigging it is slightly easier, when the arms are in a relaxed, around 90° pose with slightly bent elbows. For mocap capable characters, it is better to use 90°, as far as i know.
But that's just me and I might be wrong, let me know if thats the case
edit: what i mean is this (sorry for the bad liquify
I'm gonna go ahead and say that your Motionbuilder statement is a bit incorrect.
Motionbuilder will let you do whatever you want.
Your results may vary depending on what you want to do though :poly136:
At work we have two poses for the character. One 90 T-Pose and on 45ish V-Pose.
Both of these are in the scene, in their own take ("clip" if you speak Mayaish)
I've worked with different motioncapture sources, from our own mocap suit, to external stuios' solutions. For our own suit, the software associated with it comes with a fbx exporter, and all an animator has to do really is to just load the exported animation onto the character in Motionbuilder.
Has nothing to do with whatever T-Pose or V-Pose he might be in when you do this.
Also worked with pure skeletal data, which had the same namestandard as we're using on our character. All an animation has to do then is to just merge the skeletal data file and let it overwrite the characters joints basically.
Again, doesn't care about T/V-Poses.
However. Working at home with some free mocap and a Call of Duty 4 SDK rig that was set up in that V-Pose Hugging kinda bonanza (arms not straight down, but kinda hugging forward) and the mocap not starting in a similar pose, gave me shitty results. So what I had to do was to add one keyframe in the raw data that matched the freaky CoD4 pose.
All in all, I don't agree that MB wouldn't allow you to retarget data because your character is not in a 90degree T-Pose a'la nazi.
You may have different results if your shit is all messy, but MB will just eyeball you and let you mess your shit up
But if you do it correctly, it will still give you the ol' evil-eye, but your shit will rock
Typically a shoulder is compacted when in the T pose (90 degrees) and fully stretched when the arms at the sides (0 degrees). So 45 degrees is the mid point.
What we normally do is build in T and just before taking in to a sculpting app we move the arms to 45 and blend the knees and elbows slightly so while sculpting care can be given to the actual elbow and knee nubs which can easily be glossed over or forgotten in the clothing. From there we rig in this pose and if needed we will deform the skeleton back into a T pose if needed for things like mo-cap. The rig pose is always independent of the mo-cap pose and there isn't a need for mo-cap to force the modeler to work a certain way, that's just nonsense.
Q: If that's the case then why are legs straight the knees are compacted!?
A: Knees are a single axis joint they aren't ball joint sockets like hips or shoulders so its much easier to predict their deformation. Because of this it isn't necessary to impose strange restrictions on a modeler like working off axis at 45 degrees.
Because hips, shoulders and wrists have such a wide range of motion along multiple axes, a lot of care is given so they have the best deformation possible.
The thing is, I recently loaded a character in a 45° pose into MB 2012 and when i started characterizing the skeleton, it would give me a warning: "ArmUp (or whatever it was called) is not horizontal to the X axis" and it wouldn't let me proceed. And when i put a t-pose char in, all worked without a problem (the source file was starting in t-pose as well). So, thats how I came to that conclusion.
But I'm also sure now (after reading your post :poly136:) that it can't be as extreme as i stated it to be. I was just not aware, that you could just go ahead and ignore that horizontal on X warning. Then again, why wouldn't a software let you do something like that?
Brilliant! I never thought of that. Will try that out next time, thanks!
But if you have some spare time, maybe you could explain this to me:
So, what do you use the V-Pose character and what do you use the T-Pose one for?
Also, do you use two rigs? One for keyframe and one for mocap? (if you are allowed to say
edit (you probably posted, while i was writing my post:)) :
Mark, I see your point.. A V-Pose gives you a between-two-extremes-pose, so i guess it makes more sense to use that one.. Also, about the mocap pose, we don't use mocap where i work, so i just assumed that it would be modelled in a way that accomodates to the fact that mocap will be used. Of course you can always deform the skeleton back, but i didn't think it was a valid method. Thanks for explaining!
That's kind of a seperate issue from the particular bind pose you choose to use.
I'm guessing "ArmUp" is your bicep joint?
Didn't the program refer to the character not facing Z, when it said that the arms are not aligned with X?.
I mean, was the character perhaps facing X (or God forbid, Y:poly136:) with its arms on the Z plane?
And that the t-pose character that you loaded in later (assuming its a different character?) was facing the correct Z direction?:)
No we use the standard Motionbuilder character rig for both keyframe and mocap cleaning in the same scene.
I don't understand what you mean with rigs with separate joints, constraints and such that can't be loaded into MB. If you are talking about AUX effectors and stuff, then yeah, they tend to be left behind when merging a character into a scene.. I'm not tech enough to really say any of this of the top of my head though.. :poly127:
But separate joints is more than fine in MB, just use Character Extentions and add the extra joints there..then they'll be a part of the character and animation done on them can be loaded over multiple files.
I just asked my TA and he gave me an answer which makes everything that I've stated sound like a bunch of shit... :poly141::poly142:
I asked him what our T and V poses are for.. he said that T is for characterization (but again, we are facing Z properly, and if that was the problem you had then it still doesn't matter if its T or V) and V is for the modelers.
I've never used the T-pose as a base for animation though, and always go from the V pose as its more relaxed and just makes more sense to me :P
Also a lot of the rigging takes place (at least in 3dsmax) on the joints local axes so it can flipped upside down and rotated 354.001 degrees in world space but in local its still relative. But even with it relative its still easier to stick to the main axes while setting things up and there are a few world space options for rigs that its just easier if things aren't all deformed and weird.
One other drawback of the T-Pose is that it makes it hard for the modeler to define the shoulders and it often gets left up to the skin weighting to define the shoulder. In such a hard to weight area it can be a bit of a nightmare balancing it all, which becomes more troublesome if the modeler didn't accurately predict how the shoulders will deform. Which leads to a bit of rework and fine tuning that might of otherwise been avoided with the arms down in a more natural pose.
I also agree with vargatom it's hard to judge proper arm proportion when them in a T-Pose. arms often end up longer than they should be, and hands larger. It's much easier to imagine where the fingertips would come to rest on their leg if the arms are at 45 degrees.
A while back I had created a simple shoulder rig for our modeler to drop in his scene and skin wrap his mesh to and then he could deform the shoulder and move the arm around to see how the shoulder would possibly deform. This helped him sort of learn to predict shoulder deformation and he got really good at creating awesome shoulders really quickly.
Arms down 45 degrees.
Elbows bent 45 degrees.
Fingers half curled.
Legs spread.
Squatted down to bend the knees.
This gives the modeler the most control over how the character will deform. It takes a little longer to rig, but it's worth it.
Palpatine pose because its like the emperor sitting down?
Yes the character was facing the z axis. I had seen that in a tutorial at that point, that said that it was crucial for the skeletons to face z, so I made sure, the character would do so.
I meant custom rigs, that have stuff like stretchy leg joints and don't necessarily have a continuous joint hierarchy (leg roots are constraint to the main root, but aren't parented, for instance). I tried importing these kind of rigs into MB but it would not recognize connections like that. (maybe I didn't import it right, of course) :poly136:
What I'm confused about is: What if you wanted to apply mocap animation in addition to a char that was animated in maya with such a rig? For instance, a run animation in maya and then applying a jump mocap animation following the run. I'm guessing you would create a rig that is MB friendly in the first place, without all the above mentioned fancy shenanigans?
never really seen any custom rigs like that.. if you use motionbuilder, then you use their rig or fuck of, kinda :poly141:
yeah.. well, if you're doing a game where you want animations to be stretchy and cartoony and what not, then you probably wouldn't use motionbuilder in the first place.
but yeah, you'd have to go MB friendly..
I mean, you could probably get some FK stretchyness going on (and then if you're tech enough, probably get some IK driving it as well) if you have one hierarchy that is MB friendly, but that is not really skinned to the character, but has joints attached to it which are..
so, you'd have your right leg with 3 joints, that MB can rig with it's rig..
then you'd have one joint parented to each one of those, which is the joint that actually has any skin influence.. and then just use them to stretch the mesh I guess ;D
It'd get fucked up fast though, having rotations offseted and what not..
meh I dunno...:poly136:
Thanks guys for explaining this to me! I've often found myself wondering how that worked and since I couldn't find articles or tutorials or whatever explaining that kind of workflow, the question remained unanswered. So now I have a general idea of this type of thing
Thanks again guys!