To T Pose or Not to T Pose

splicer

This is my question. Should I model my characters in a t pose or not? The one of the right is my original model, I later changed it to a t pose so I could make use of Maya's Human IK quick rig. This is good for using premade animations in Unity, however I am sure there are down sides.

Would love some opinions on the matter!

Mark Dygert

A pose, it lets the modeler define the shoulders and they won't stretch that bad when the arms come down to a natural pose. But really, its a minimal choice and not something that would make or break a pipeline.

Alex_J

I have been using the A-pose, and my major issue is that the initial bind grabs a lot of the torso with the shoulder joint in the armpit area. Not a big deal as once I've fixed the issue once; I can usually copy my weight from model to model. But it does mean an extra 10-30 minutes of fixing up that armpit area each time. 

By the way, you don't have to use the t-pose with humanIK. I am using humanIK and use the A-pose. I think that would only matter for the controller setup? Though I'm not sure. I am using pre-made animations, so I don't need any controllers.

musashidan

+1 for A-pose. Much more natural and easier to work with.

splicer

Thank you for the feedback!
@Mark Dygert I am noticing this too as I sculpt, the deltoid pushes up against the collar bone naturally, does make it problematic.

@BIGTIMEMASTER I think I see what you mean. I am using their controller and it is breaking the initial Maya IK when I move the bones down. I'm going to do some tests with the original model and see if I can get it to work! Thank you for informing me of this!

@musashidan I appreciate the input, I will change on my new character. She was A pose, then I switched to T, will go back to the A pose!


How she started.

How I've been working on her, hasn't matter since been doing hard surface, but going back to body now, so I will revert back to A

splicer

So I fixed her T pose to an A pose, but I did not put the arms down as much, due to the arm put comment made by you @BIGTIMEMASTER.

Alex_J

Keep in mind, my comment about the initial bind not reading the armpit area well is only a minor inconvenience thing that matters if you are doing a lot of rigging on similar characters. In my case, I am making a lot of modular characters that are more or less shaped the same, so I can just copy and paste the weights from one to the other to save the bulk of work. 

If you are rigging one higher quality character, it may make more sense to choose your pose based on what makes your sculpting the easiest/gives the best result. 

musashidan

BIGTIMEMASTER said:

If you are rigging one higher quality character, it may make more sense to choose your pose based on what makes your sculpting the easiest/gives the best result. 

This is why learning to properly use layers in Zbrush is important. A lot of people have no idea the workflow power that layers can give you.

splicer

@BIGTIMEMASTER I do see your point now, makes sense now. I like this arm placement now, the T pose felt very stiff on her, and the shoulder area was not working. I'm going to stick this one out and rig without the controller in Human IK Maya, see how it works out.

@musashidan This is something, I am not familiar with, using the layers. What are some of the advantages? I did see a presentation at the Zbrush summit a few years ago by an artist for the Far Lands game, showing the facial expressions baked out using layers. Was very cool! I would like to know more examples though, anything to increase my efficiency and do things the best way possible, I will learn no problem.

Alex_J

Same concept as layers in photoshop or Substance Painter. Allows non-destructive look-dev while you develop.

musashidan

@splicer The main benefit is, of course, working non-destructively. An example of this would be sculpting a head. You would sculpt your forms up to primary and secondary shapes. Then add layers for tertiary forms: general pore layer, directional pore layer, eye wrinkles, etc. On top of this you can add an asymmetry layer. The advantage being that you can turn this layer off and continue to work on primary/secondary forms using symmetry, while maintaining your asymmetrical layer. Always adding layers at the highest subdivision level and making use of the power of being able to make bigger, non-destructive changes to the lower layers and have the detail propogate through your levels/layers.

This is also true of creating further expression/phoneme/morph or blend shape layers.

Another example is creating layers with the eyes fully open/closed. This can then be used for unwrapping/texturing/rigging

And in your case you can create 3 or 4 layers transitioning from A-pose to T-pose(and beyond) and switch between any one of them for sculpting. Then switching out again for rigging. It is very flexible and - again - non-destructive.

A feature of Zbrush layers that a lot of people don't even know exists is that you can even import a saved version of your tool directly into a layer.

Alex_J

musashidan said:

Another example is creating layers with the eyes fully open/closed. This can then be used for unwrapping/texturing/rigging

And in your case you can create 3 or 4 layers transitioning from A-pose to T-pose(and beyond) and switch between any one of them for sculpting. Then switching out again for rigging. It is very flexible and - again - non-destructive.

Can you recommend any tutorials that show how to do this? (even if it's your own)

I think movign a character through poses early on could be very powerful -- maybe essential -- for game characters. Working on my first game project and starting to see my characters in action, one thing I am learning is how different they look actually moving around in a game environment versus in a static pose in Zbrush or Substance Painter viewport. Once they become alive, I notice all sorts of things I missed before. I think it would be extremely useful then to bend and stretch a character around as early as possible, so that I can start to imagine how it's going to look as a final in-game product. 

musashidan

@BIGTIMEMASTER sorry mate, I don't know of any. It's something I worked out by experimenting myself. You basically just create layers and pose your character. Then you just enable/disable layers or blend them with the sliders.

Alex_J

musashidan said:

@BIGTIMEMASTER sorry mate, I don't know of any. It's something I worked out by experimenting myself. You basically just create layers and pose your character. Then you just enable/disable layers or blend them with the sliders.

Thanks. It is something I have known is possible but haven't got into myself yet. Figured would be nice to get  a jump start by having someone show me step by step. I'll take a look around in the near future and will share if I find anything. 

RN

I think ZBrush layers are similar to Maya's blendshapes: http://docs.pixologic.com/user-guide/zbrush-plugins/maya-blend-shapes/

So you can do this it directly in Maya if you prefer. You duplicate your mesh, model it (move the arms down etc.) then add it as a blendshape target on the original mesh (or whatever the workflow is for doing that in Maya).

splicer

musashidan said:

@splicer The main benefit is, of course, working non-destructively. An example of this would be sculpting a head. You would sculpt your forms up to primary and secondary shapes. Then add layers for tertiary forms: general pore layer, directional pore layer, eye wrinkles, etc. On top of this you can add an asymmetry layer. The advantage being that you can turn this layer off and continue to work on primary/secondary forms using symmetry, while maintaining your asymmetrical layer. Always adding layers at the highest subdivision level and making use of the power of being able to make bigger, non-destructive changes to the lower layers and have the detail propogate through your levels/layers.

This is also true of creating further expression/phoneme/morph or blend shape layers.

Another example is creating layers with the eyes fully open/closed. This can then be used for unwrapping/texturing/rigging

And in your case you can create 3 or 4 layers transitioning from A-pose to T-pose(and beyond) and switch between any one of them for sculpting. Then switching out again for rigging. It is very flexible and - again - non-destructive.

A feature of Zbrush layers that a lot of people don't even know exists is that you can even import a saved version of your tool directly into a layer.

What you are saying is something I am looking for right now, as I begin to detail out this character, this would be perfect. I will experiment with them and see just how they work. Will probably wait till I start on another project, just because this late in the highpoly, I worry about errors.

melviso

Nice work @splicer
Em… what of during mocap, Isn't the mocap life actor in T-pose before recording motion or A- pose is now used? I have always thought u had to have characters in T-pose if there are to be used for mocap. 
If hand animated, A pose is more suitable.

musashidan

@melviso If you look at the UE4 mannequin character, it is in A-Pose. The arm pose in that case has to be matched for retargeting(although this can be done after the fact in-engine)

melviso

@musashidan   That's true. I completely forgot about that. 

Mark Dygert

Yea, you can have a few different poses at different points in the pipeline, use whatever works best for you and your workflow.

Quite a few mocap systems use a T pose for a common, default calibration pose. I've worked with two that had options for A pose also. That doesn't mean you need to model in that pose, just that your rigged up skeleton and your actor need to move to that pose for a few frames so things sync up.

When modeling and even rigging, it can be easier to work in a T pose, but at some point before skinning or even during or afterward you move it to a more neutral default pose and fix any issues that might have popped up, IF there are any. It's not really going to break anything, its all about whatever is easier for you at that point. There are a half dozen ways to translate the pieces in various apps and to account for differences that it's not really an issue.

splicer

Some great information going on here. Now I'm looking forward to trying to rerigg my first character,  good as the next one is more complex.

Also, the first one, I did not model the inside of her mouth, to keep it simple, but her eyes are. Sooo, going to try a simple face rig too.

Atleast have her eyes rigged so I can make her look around while in Unity Engine.