T Pose
polycounter lvl 18
Just curious on what some of the T poses you guy's use for your basemesh. As we have been going over what might be best for a relaxed pose and for animation purposes. At the moment the character is relaxed, with bent knees and a half lowered arm and bent elbow towards the front. The hands are slightly drapped and relaxed so that its about 50% inbetween positions for texturing and such. Preciate you guy's sharing!
Spark
Spark
Replies
Though I feel like the pose that I keep seeing at the top bar of Perna's Jazz hand giving soldiers seems like it might be a good type of pose for a model.
Before sculpting a quick rig is done:
- Arms relaxed to a 45 degree angle (T-pose being 90, arms at the side being 0)
- Forearms bent 15-20 degrees
- Legs parted 20 degrees with a slight bend at the knees
This pose caries through the rest of the process.
Most of the time the person doing the rigging/animation will clean up the default pose after the character is placed in the environment (seated, standing, pacing ect). Most of our animation takes place in conversational dialog not the traditional run and gun anims.
http://www.3pointstudios.com/portfolio_characters.html
I think he discussed some of the reasons why, a search might turn it up.
i think.
I know it is much easier to model and skin a mesh (particularly fingers) when it is straight, but the straighter the form, the more collapsing you'll get in deformation.
I guess it also boils down to what the character will be doing the most, driving, shooting, jumping....
I'd be interested in seeing how the characters for fighting games (Soul Calibur, Street Fighter, etc...) are rigged/modeled as they all have very extreme poses and ranges yet still seem to hold up well. Anyone have info to share there?
That second part is not really true - you're right that it's easier to rig/skin a hand for example when the fingers are straight out (plus it gives you more consistent joint orientations, which can be useful) - but I don't think that just because the form is straighter means you'll get more collapsing. It really depends how you model the mesh around the joints, and how you position your bones, and how you set the skin weighting.
Fingers in any pose can be moved to any other pose without nasty collapsing, it's just a matter of knowing where the polys need to go relative to the joints, and what skin weights to apply.
Per's models are a pretty good example of this, you can sort of see how the arms are posed in a way that is about halfway through the range of motion.
For hands i dont think its really that big of a deal to curve the fingers on characters. Unless you've got a FPV hand, then starting with a more relaxed(closer to a gun grip post) is best, that way your UVs again are set up optimally, not constantly being stretched as the hand is deformed.
If the arms are at 0 degrees when resting at the side, and at 90 degrees when in the T-Pose, 45 degrees is dead center.
It's also important to remember that arms rotate in more then those two directions they also go forward and back (welcome to shoulder rigging hell). Rotate a T-Pose'ed model arms down to 0 and then forward 90 and you probably have issues. If its a 3rd person pers, you have your user, staring right at a big patch of ugly...
As EQ pointed out, taking the UV's into account is a good idea. Stretching the UV's so they are squished in the T-Pose can help, that way as they stretch they become more normal. But really there is only so much planning you can do, and certain amount of clean up isn't a horrible thing. Especially if someone is going to spend hrs obsessing about problems that may never materialize.
Of course optimal geo layout and bone position is a given, and a required for proper deformation, my point is, when is "bad" deformation most noticeable? i have personally found that it's easier to transition from a relaxed, or "gun grip" pose to outstretched( still talking about hands) rather than vise versa, and with good layout and bone position transition to a clenched fist. just my experience.
Haha. I just didn't think that anyone would get the joke when I re-read it this morning.
I always model the arms 45 degrees (exactly) down, it makes rigging slightly more of a hassle but the tradeoff is worth it in regards to deformation gain. Having the elbow bent is slightly more frustrating, I personally keep it straight. Then I have the hand parallel to the forearms and all fingers straight, since that'll give you the easiest access to all poses.
The biggest hassle by far when it comes to rigging a character's arms is when the shoulder, elbow and wrist joints aren't aligned in the front view and I'd avoid that at all cost. Apart from that it mostly depends on how advanced the rig should be. A solid twist control is better and easier to set up when the wrist is algined to the forearm, especially if the arms need to switch form FK to IK. if you want - unlikely in games though - stretch options, you need the elbow straight as well.
Bottom line is first look at the features the rig should have, that'll give you a good idea of what you should avoid to start with, Other than that it's just a tradeoff, how well can your rig deform the character and/or how well can the rig funtion if it already is. But again, make sure the shoulder, elbow and wrist are in a single straight line from the front viewport. Same for the legs, which I always model straight down, feet pointing to the front, even if the feet/shoes clip through eachother. Any other method is quite unproductive in almost all cases, apart from making the models look better in 't pose'.
I have to disagree with almost all of this. If you're worth your salt as a rigger, the modeling pose doesn't matter as far as rigging is concerned. All that really matters is for deformation, and a more relaxed pose is going to give you a better result- the focus of the basepose needs to be deformation, not rigging. If you are basing your basepose decision on rigging, you are hiding a lack of rigging ability at the cost of deformation.
The only time I've ever had an issue rigging something based on the basepose was, as you mentioned, when something is already rotated which would require a twist bone to rotate. Other than that, I can't think of a single case in practice or theory where the basepose has impacted my rigging in any significant way.
I also agree with Per and Ninjas about the pose. I've done nothing more then take the model as it is from the modeler and do a simple rig and slight pose change and had people respond much more positively...
I was wondering if there was a reason for it? Doesn't it mean that you have some twisting in the shoulder region by default when they're posed with guns held?
Cheers!
In some cases it makes rigging life (and I'd say animator life just as much) a lot easier if the model is practically oriented. More importantly as I pointed out, it makes some things (near) impossible to get right.
As an example here's a posed arm (far less extreme than yours, but technically shows the same obstacles) and the arm as it is in its default position. This particular character has the elbows straight down, I sometimes bend those a little for deformation, which does make the rigging part harder.
As you probably know, as the elbow is a hinge joint therefor the orientation of the arm is in a plane that goes through the shoulder, elbow and wrist, which is something that's automatically set when applying IK to a bone rig. Now for clean results, you'd want (need) your joints to be set up accordingly before applying the IK. If you look at how the orientation is in both, you can easily see how hard it is to set up either.
As you can see, all joints in the right chain are oriented exactly the same (including the twist bones in the upper- and forearm, which are not highlighted) so their local orientation is 0,0,0. This means a twist bone will always make a proper rotation relative to its parent, without the need to compensate for any other initial orientation. If you've ever had a twist bones snap in any way or rotate too much or too little, this is essentially the reason.
More importantly, if your rig requires the ability to swith between IK and FK, it's vital, especially for features such as twist control, that both your FK arm and your IK arm are perfectly aligned, in order to make proper switches between the two (local space vs world space based orientation). Now IK FK control, let alone stretch ability, might not be somethign most games include (although I'm guessing the former increasingly more often), but it's near impossible to get right without proper orientation.
Then there's the case of actual result. You often see hands which are posed. Fingers spread (but in a single plane) or hands posed like in the top example. As MoP pointed out earlier in this thread:
Hands are infinitely easier to set up as well as animate when you model them looking as boring as possible. Yes, it is possible to stick joints in a preposed hand, and even get the orientation of each individual bone correct, but it takes a lot of time, will not improve anything deformation (possibly make it worse) and it's not as easy to animate with. Here you can see a front view of both hands:
It should be obvious that the right version is much easier and quicker to setup. Orientation is the same fr every bone and they're all in the front view plane so they can all share the same controllers and values. Now if you have to animate with a preposed rig, you'd start with the same preposed hand of which all values are 0, in other words they do not have corresponding values relative to their orientation.
But most importantly having preposed fingers will not help the final result at all (unless of course you don't have the bones available). It's a lot quicker to model, it's a lot quicker to rig, it's a lot more functional when animating as you can always build poses from an actual base pose and the results will be better overall.
I hope I've explained properly why I think the modeling pose matters a lot and it some cases is even dictated. I said in my previous post: "Other than that it's just a tradeoff, how well can your rig deform the character and/or how well can the rig funtion if it already is." But unless you have a good reason, such as deformation especailly on areas such as the elbows, I'd always stick to the easiest solution for all. In terms of the plane of the arms and legs, you should never even consider it in my opinion, always make those face the front. It shouldn't be the job of a rigger to clean up after a bad production pipeline, especially at the cost of functionality (again, if needed). Thinking about rigging and animation before making a model is vital in my opinion. Everybody takes deformation into account when building a model, but the rigging itself is just as important for deformation and if you want some slightly advanced features, even more important.
Unposed models look really really boring, but the upshot it its a lot easier to make it look good in every step after that. I mainly rig and animate my own models and I've found this is the way to get the best final results.
We've got a rig where the finger joints are perfectly straight out and it works great, also the arms are straight in the front view (although bent at the elbow) like SkullboX said.
As I was going to say earlier but didn't - any time you have stretched UVs in a pose which your character is seen in a lot, that's not the fault of the base pose, it's either bad skin weighting or bad UV-mapping. No matter what pose you model in, you will always end up with the same amount of stretching at certain key points (shoulders, knees, elbows), so if you set up your UVs to account for this give them a little more space, then it will end up looking the same as if you had modelled it in a crazy extreme pose, except it will be much easier to rig and set up very predictably.
Also, correct me if I'm wrong, but I think out of everyone who's posted in this thread so far, SkullboX is probably the one most qualified to talk about good rigging...
Hands are not significantly more difficult to set up or animate with when in a flat pose, that is flat out wrong at worst and personal opinion at best. You are seriously talking about the difficultly of placing joints in fingers? You find that difficult? Furthermore, it won't matter with deformation on a single custom character, but modeling hands in a rest pose will give you a much better knuckle, and if multiple hands are sharing a rig, the more 'between' the pose, the better; you have much more leeway.
It isn't obvious. I don't consider placing bones the most rigorous of a rigger's tasks (well it is once everything is completely automated), it is almost negligible. Nor do animators animate with fingers numerically- saving poses doesn't matter whether it is a straight or rested pose, nor does it make a different with SDK's, or anything, really. Maybe animating with sliders? But even then, you are starting with a pose and adjusting.
This doesn't even make sense. It WILL help the final result for all the reasons people have mentioned (better deformation and closer to the most viewed poses). It isn't quicker to model, it is marginally quicker to rig, and the difference to the animator is non-existent- I have never once looked over and seen an animator using the basepose.
Something is wrong with your production pipeline. And furthermore, you are making problems with it. Let me explain:
First, you have stated numerous issues that I, nor anyone I know, have never run into, despite being in the same exact circumstances required to experience such an issue many, many times. Many of the issues you have described simply do not exist if you are doing a good job rigging, plain and simple. Thank you for elaborating in your post, so I can be sure this isn't a communication or assumption problem, it is a production problem. If you are having these issues, it is time to update your rigging and scripting practices.
Second, as I previously stated, you are sacrificing deformation for ease of placing bones (I will not say ease of rigging since I wholeheartedly disagree it makes it any easier to rig). One of the few things that cannot be fudged is skinning- as someone who has skinned an ungodly amount of objects of all sorts, I can say with certainty you cannot fudge skinning- it is one of those things that cannot be compensated for, because it is the very end of the geometry pipe, there is nothing after it. This is especially important to realize, since with a good pipeline, no matter what the basepose is it shouldn't make a difference to rigging or animation- it should only matter for deformation. Weighting is commonly used to try to hide or overcome problems because it IS at the end of the pipe- this is really the complete opposite of what a good pipeline is and what good deformations are made of. Someone made a joke about using a posed basepose because the mesh was snapping to its bindpose ingame- it was funny but what you are doing is really the same thing. You are making decisions based on problems that should not exist, based on bugs. You are making a pointless sacrifice to deformation, because a pipeline and rigger is not up to snuff.
That isn't true no matter of what pose you use- if your animators aren't working with a library of poses, or some other way to work with fingers (and other parts), your pipeline is bad. Almost any time you are relying on turning on angle snap and some combination of selection and rotation to achieve some key pose, your pipeline is not taking something into account. This is a terrible reason to believe something or agree with someone. Our field moves WAY too quickly for experience to matter in a technical sense. I will be the first person to give reverence to experience when it comes to something that experience can inform- but these are almost never technical issues and it is sure as hell never the case when it comes to production pipelines or their technical aspects.
Now we are using UV's to overcome a bad basepose? Tsk tsk. If things are modeled in a relaxed pose- as many in this thread have discussed- you minimize deformation problems, while requiring no extra work. Why do more work than is required?
And lastly: I'm not saying everyone's production pipeline needs to have all these things or it sucks. I'm saying what they should have, what there are no excuses to not have in the works. There are lots of problems with our pipeline at work, though a fraction from when I started here. But the theories I'm talking about are solid; nay, more than solid I think. And that is what this is about, theory as informed by experience. If SWTOR is the best looking game ever it'll still have mediocre human deformation, it was still a mistake to use the T pose, and Max sucks for rigging and animating. All our pipelines have problems and will always improve- however, what I don't think it is acceptable to do is to sit on our heels and not realize that things may be very different from what used to work for us.
I believe that to some extent, the attitude of a given bindpose sometimes appears in the final animated model. Yet when done wrong, 'posed' bindposes can bite you in the back - like overly bended knees.
I personally enjoy a 45 degrees down arm position, but a straight T is not that annoying IF the forearms are done correctly
I often see straight tposes done with the palm of the hands facing down - I try to avoid that because the straight look of the T suggest an all-purpose 'rigid' rig, yet palms facing down means that the polygon flow of the forearm needs to twist on its way down to the wrist (since ulna and radius are not paralel in that case http://www.noelkingsley.com/blog/A4ulna.jpg), wich is more of a special purpose 'twisted' rig situation.
So yeah I think that Per's JazzHands pose is great in that regard.
---
Oh Btw Rob : I think all your points are very valid, but there is also some deep-rooted issues in our industry making it very likely for the roadblocks Skullbox pointed out to appear.
First off, there is almost NO game studio out there willing to share it's pipeline/developpement tools after the projects ships. This is in my opnion the biggest problem here, since it forces almost every team to start from scratch and hit the same issues again. One could argue that since the tools took time to build, they should be kept secret. This is bullshit, since the whole idea of sharing is that if you share, people share back. I am forever grateful for the Blurscripts to be out, as well as the Criptic rig. And look at Maya : it's a great, powerful collection of functions and an open tool, but it is barely usable out of the box for stuff like weight transfer, or even pose sharing from one file to the other. All these tools are to be created as proprietary tech by every studio using the app, and I think that in 2009 it shouldn't be the case anymore.
Second, even if this industry is growing fast, many things drag behind. Sometimes because of bad decisions, sometimes just because an old yet solid enough system can still be used efficiently hence avoiding extra developpement time. Like, the 10 years old CharacterStudio rig still use in Unreal3. It's old, outdated, but solid enough so no need to go fancier.
Lastly : Skullbox can rig but is first and foremost a fantastic animator. (Magdalenaaaaa!) Even if the issues he pointed could very well be avoided by doing things you mentionned (mostly, stuff relying on very solid custom tools), many smaller studios without dedicated riggers or tool guys cannot afford all that and need to fallback on more basic techniques and pipelines especially when the rigging and animating are taken care of by only one person - which is I think what SB is referring to...
Great thread!!
I certainly agree with you here, but I guess I've personally found the best way for me to get the most consistent results in deformation has been using a straightforward approach to the models orientation (again, the example above was a very simplified one). But that was indeed probably in part due to my inability to produce the same amount of control in the rigs that were posed. I guess it wasn't really my place to comment there to be honest.
At least it's cleared up.
I agree 5000% and work every day towards improving this. I am totally focused on getting more sharing and distribution out of bioware and ea. It is a very slow process but wheels have slowly been turning. Unfortunately, it isn't the corporate people who are the biggest roadblock- perhaps they will be, but they haven't been an issue. It is a developer mentality, and we can work towards changing it. But that is for another thread.
Now I disagree. CS is NOT solid enough. Max is not solid enough to build a pipeline for. Maya is going to fall apart sooner or later. Like you said, these decisions can cause the mentioned roadblocks to come up- how in the world does that make them good enough? They aren't, it is legacy, it is being afraid to learn and try new things. It is ignorance of potential. It isn't saving development time, not even in the short term. This isn't about switching programs or anything like that- it is a much more significant change of pipeline paradigm from what most, but not all, of the industry still use and cling to.
This sort of reminds me of the P&P discussion last week about baking in diffuse lighting- systems are now at a point where very minimal diffuse lighting needs to be baked, if any, but some people bake it anyway out of habit or outdated ideas about what is needed. What I'm describing regarding rigging doesn't require fancy stuff, and if your sacrificing deformation for rigging, you should look at other options (such as using something off the shelf, there are many great options for all major programs). Obviously what you do is your own business, but as I mentioned in my previous post- we should be looking forward with these discussions, not backwards.
What matters- and believe me it is hard to say this as a nuts-and-bolts minded person- is the result in the end. If you are making great animation, far be it from me to impose what I think where I have no business. But like I think we agree now, we need to be careful when we advise people- we don't want to restrict discussion to self-declared authorities, but nor do we want bad information clung to and defended instead of being debunked. In that case, I am really glad this thread has maintained a civil tone and you handled my dickishness with far more civility than I could muster or I'm used to on these forums. There's nothing more someone can ask for, so thanks.
we found it's hard to judge a model if it comes in some wind-up-toy like pose with straigthened fingers and legs.
pipeline these days is one hell of a best to deal with, i sometimes want back the times when we only had to deal with character studio and some primitive skinning.
btw. i got to look at a number of folios in recent weeks - funny how many artists are shopping work around that has clearly the wrong proportions. at a certain well known project we got a bunch of folios from they seemed to have established a mannequin with way too short legs and long arms. always entertaining to look at but maybe not what you'd want as a starting point for everybody's work.
Why make rigging and animation hard and possibly lead to worse deformation? I see a posed hand as being a purely selfish maneuver by the modeler based on aesthetics. Why make it harder on the next guy, especially when hand deformation is easy enough to control with straight fingers.
Why I see it as selfish: Finger joints are not shoulders.
Finger joints rotate on one axis (with perhaps some rotation slightly on another). Deformation on one axis is easy to plan for and tackle, so at that point, deformation isn't an issue and its worth it to make rigging and animation easier.
It makes sense to pose shoulders because there is more then one axis of rotation that is going to be pushing and pulling on the shoulder verts. Rotate a T-pose down to 0 and then forward 90 (kind of like this), you're probably looking at a patch of ugly, especially if the player POV is 3rd person. Fingers never bend 90 degrees on two axises and because of this they don't need to be posed.
The shoulder deformation issue actually should be tackled properly with a more complex shoulder rig, using joints to mimic the shoulder blade, so if people want to create and deliver models in the T-Pose they can, it would also make rigging easier. But because we work on games and smart people with very little understanding of anatomy make the calls on what bones are "needed", we end up in this mess of forcing the modeler to make up for a skeletal deficiency. Like Rob pointed out, maybe we don't need to hold to the old ways of doing things and its time to reevaluate what bones are needed? Helper and twist bones are becoming more prevalent, but I think we need to sit down and redesign the start of the arm chain from scratch?
Based on my personal and professional experience the T-Arm Pose makes it harder to gauge arm proportions, just like it can be hard to gauge leg length if the knees are bent, or rotated away from each other.
I gauge arm length where the finger tips fall on the thigh. I encourage our modelers to check this with a quick rig/rotate and continue to model in the T until its ready for rigging, then the arms are posed and cleaned up at 45. I created a very simple rig they can drop in, apply skinwrap and rotate to check arm length.
In short, yes we can rig complexly posed models just like modelers could be forced to model using this, but why?
Modeling a hand in a natural position makes it much easier to correctly judge whether something's good or not. A modeler runs the risk of becoming blind to flaws in a model after working on it for a while, ofcourse, but I find it easier when it's at least in a natural pose.
So in short, we can model a human in the pose of a robot, but it makes it harder.
If they are doing it for proportions then they can do a simple rig/bend test.
The biggest positive impact on having a posed hand is pre-rigged eye candy. If you have some advantages I'm unaware of I'd love to hear them, but know it creates slightly more work and isn't necessary.
Modeler: "If I pose the hand it makes deformation easier, I'm helping!!"
Rigger: "Deformation isn't an issue, it actually could make it worse..."
Modeler: "But it helps me get my proportions right!"
Rigger: "whatever you have to tell yourself to sleep at night. By-the-way I'll be spending tonight rigging your hands..."
It's 7pm everyone has gone home the rigger finds out the proportions of the hands are actually off
Rigger: son of a b!tch...
I guess the answer to the question at hand is, "whatever works best for your team, keep an open dialog and don't assume that what you're doing is great for everyone else".
Seems the ultimate solution may be to jes admit that skinning is an act of sculpting as much as any edge looping. Which includes skeleton placement and orientation.
been experimenting with boning at the earliest stages of sculpting ( scripting automated binding, unbinding and re-transfering weights to facilitate skinning as sculpting. cuz manually doing so iz tedious )
seems to be the "right" direction.
Aren't Riggers jes the guys who can't model or animate anyhow? ( kidding )
edit: Use to use the same werkflow years ago in Mirai. If I remember right, bones behaved intuitively even during topology tweaks. Fer shits and giggles I installed Mirai on vista x86 and lo and behold it's werking!
but seriously, we don't really have any issues with T-pose stuff, we do a relaxed pose, not for the artists, or the riggers, but for the animators, they want the the model in the most relaxed pose when then start animating.
the biggest argument for a T pose is motion capture as they usually calbrate the rig with a t pose before they capture the movement.
but this is really a non issue, do what works for team. as a modeler i have no problem with a relaxed pose or a robot pose, as most modelers should. if you get paid to make stuff there shouldn't be arbitrary circumstances that make it so you cant do your job. i could just see the look on an art directors face if you were like, "i cant model that cause the bones are straight, i could make it look better if we change the whole pipeline right now just for little ol me"... really, well find a new job. part of working on a team is compromise and it takes me just as long to model a straight limb as it does a bent one, so i could really care less as it impacts me 0%
Obviously if I was a modeler on a team, I would model whatever the concept artist output, even if it was lame, and I would do what the AD told me to do even if it made no sense.
Spark
Now I'm primarily a rigger who sometimes models.
And in both instances my preference for bind pose is slightly bent knees. Arms down between 20-60 degrees. Slight bend in the elbow. Hands following the forearm direction, and fingers slightly curled.
I passionately hate the T-pose with the arms up at 90 degrees. It makes my life hard as a rigger, and as a modeller.
One problem I see with T Poses a lot is short arms. Or else no bend in the elbow, which makes setting up IK arms such a pain for riggers.
Our automated rig will automatically rotate the arms up into a T pose as soon as you build the rig. That's for copying animation and for code because our engine idiotically exports bones with world space orientation instead of the joint orientation you set in Maya...
Any custom rigs I build I keep the arms down around the 30-45 degree angle. 98% of the time that's where the animator will be moving it as soon as they start animating anyway...
On the TPose shoulder issue. It sucked a little back when I was only painting diffuse maps because you would have to account for the relax in the UV's for when the arm was down (which is the most consistently seen in-game position). But it sucks donkey dong now that you have to generate normals from a baked model. 45 degree angle is the way to go.
- BoBo
The rigger next to me wants the hands to be parallel to the flor in order to have the Y axis pointing up so the animator are alright with it (meaning that they have nice curves and controlers are oriented correctly). So, that means that we can have arms bent at 45 degrees but we have to have the hands at 0 no matter what.
What do you think about that ?
In most cases you can create rigs that align in one direction, that way they are easy to create, such as arms flat in one direction like a T-Pose is easier to "rig" constraints and things rather than arms at 45 degrees perfectly aligned to a mesh. Later after rigging is complete then you can fit the skeleton to the mesh and then bind it.
So you might have a rig pose, a skin pose and maybe even a mo-cap pose and none of them need to be the same.