Tag Archive: IK


An IK Leg and a Bug Hunt

Im not sure I can truely put into words how frustrating creating the IK leg for this rig was. I do know, however, that creating a stretchy IK leg should not take almost 48 hours to get working.

Before I duplicated my bones, I decided I wanted to put some extra bones along the lower leg to enable smoother twisting. To ensure things worked nicely later on I wanted to ensure that the three new bones I inserted were exactly a quarter, half and three-quarters of the way along the calf. To do this, I parent constrained each of the three bones to both the knee and the ankle, and made sure maintain offset was unticked. This placed all three bones exactly half way between the knee and ankle. I then changed the weighting of the parenting to two of them. Femur01 was weighted 0.75 to the knee and 0.25 to the ankle, whilst Femur03 was weighted 0.25 to the knee and 0.75 to the ankle. I then deleted the constraints and used comet tools to orient the joints correctly.

Joint Creation 11

Once this was complete I created two duplicates (one for IK, one for FK). I then created a third duplicate in which I deleted the three femur joints and reparented the ankle straight to the knee. This was so that I had just two bones in the IK system. I called it the IKGuide. I created an IK on the joints and made a simple cube controller for the foot. I parented the IK handle to this control. I then orient constrained the ankle bone to this controller so that it would not rotate as the leg moved. Finally, I created a simple circle controller and parent constrained the hip bone at the top of the IK to the circle.

IKLeg01

With the IK built and working, I started to set up the IK stretch. Like the IK spine, I needed to know how long the leg was at any one point in time. However, I only needed the straight line distance from hip to ankle. I used the distance tool for this. I aligned one locator with the hip and one with the ankle and then parented them to the corresponding controllers. Now, as the leg moved, the distance tool would always give the distance from hip to ankle. At this point I also realised I had not created a knee controller. For this I used an arrow shape. I point constrained the arrow to both the hip and ankle (with maintain offset unticked). This placed the arrow on the plane between the two, directly in the centre. I then used an aim constraint to ensure the arrow was pointing directly at the knee. After deleting both constraints I moved the arrow in front of the knee and set up a pole vector constraint on the IK.

IKLeg02

To wire up the stretchy leg I used another multiply divide node and, like the spine, the distance was wired to the first input. The second input needed to be the length of the thigh bone plus the length of the calf bone. As the joints had been oriented correctly this could be found simply by adding the x transform of the knee and ankle together. I changed the node to divide and wired the output to the “true” output of a condition node. The “false” output was left as 1. The condition node also had the distance in the first input and the length of the two bones as the second. It then compared the two lengths, and if the distance was greater than the bone length, the condition was true. I then wired this condition node to the x scale (length) of the thigh and calf. The two bones of the IK scaled nicely. Unfortunately, the ankle and foot bones were being scaled strangely when the leg stretched, despite not being wired to the condition node. I even checked the scale of both and x, y and z were all still showing as 1. This meant the bones shouldnt have been scaling at all.

FootScaleBug FootScaleBug02

I tried deleting the IK and remaking it, but the problem persisted. I then tried moving the controls with no IK present at all, and the ankle and foot continued stretching strangely. I could only assume there was something strange with the bones, so I deleted the IKGuide joints and re-created them. I set everything back up, re-wired the thigh and calf x-scale to the condition node and tested it again. I had exactly the same problem all over again. I tried re-creating the bones once more that evening but with no success. I finally decided the only option was to go to bed and look at it with a fresh mind in the morning.

As is often the way with re-visiting a problem the next day, I tracked the issue down quite quickly. I had all the joints in the hypershade to make sure my ankle and foot definitely hadnt managed to end up wired to anything and I realised there was no line showing the parenting of ankle to knee. I un-parented the ankle bone and re-parented it to the knee and the problem disappeared. I was delighted, until I found yet another problem. Whilst the ankle was no longer scaling strangely, it still was not doing what I expected when I moved the hip too far away. Despite being orient constrained the foot controller (and as such theoretically unable to rotate by itself) when I moved the hip controller forwards or backwards so that the leg stretched, my ankle would rotate.

FootBugRotate FootBugRotate02

I decided to fix the problem by simply creating a new version of the ankle and foot bone. I simply point constrained the new ankle joint to the one on the IK leg and orient constrained it to the controller again. Success! Problem solved, just not as tidily as I would have liked. It also left me feeling frustrated because I wanted to know why the problem had occured so I could avoid it in the future. Still, at least the problem was gone and I could get on with parent constraining the IK joints to the IKGuide joints. I unticked maintain offset and parent constrained all the joints to their respective guide joints. I parented the three femur joints in the same way I created them; by parenting to both knee and ankle and then editting the weights. However, what I hadn’t thought of was that a parent constraint would cause the joints to rotate out of alignment due to the ankle’s orientation. I pondered the problem for a while and decided I would simply ensure to maintain offsets when constraining the deform joints to the IK joints.

IKLeg03

I then set to work creating a control system for the toes so that I could create a simple set of foot roll controllers. Initially I decided to place a circle controller around each joint of the toes, but it quickly became clear that some of them would be hard to select.

Foot01

 

Instead, I moved the curve shapes of the controllers above each toe joint and this made them much clearer and easier to see. Finally, I also made a main controller that would be used to curl all the joints of a toe. I then began creating a simple set of foot roll controls and with some re-parenting of the IK handle my IK leg was complete.

Foot02 Foot03

Unfortunately, I quickly checked things in my orthographic side view and realised that at some point during the creation process I had managed to cause the entire IK system to move out of alignment from its starting position, despite all the controls being at 0, 0, 0.

Foot04

The only option was to yet again build the entire IK leg. I deleted all my bones, mirrored the right hand leg to the left hand side. The good thing was, that at least this time all the controllers were already built so all I needed to do was wire everything up correctly, and make sure my controllers were correctly aligned before constraining/parenting things to them. Fortunately, this time I got it right and my left IK leg was finally complete. Hooray!

The IK Spine

I finally got the mesh back yesterday, and Im happy to say its actually symmetrical this time.

Symmetrical

As such, I’ve been able to get cracking with building the control system for the toony rig. I started with the spine as I feel its such a central part of the control system, and almost everything else is parented to it in some way. My first step was to duplicate the bones that I want to apply the IK spine to. This means if I made any mistakes, moved anything by accident, I wouldn’t ruin the position/orientation etc. of the deform skeleton and I could easily delete the duplicate and start again. I tend to insert IK (or whatever is appropriate) to the name of the joint to differentiate it from the deform skeleton.

Once I’ve got my duplicate I hid the deform skeleton so that I couldn’t affect it or move it whilst working on the controls. I applied an IK spline with two spans to the spine. This means that there are control points for each end, as well as a single control point in the center to affect the curve. I created a cluster for each of the control points. These control the shape of the curve and the shape of the curve drives the position of the bones.

IKSpine01

I then needed to build the actual controllers for the spine. Comet tools provides a quick way to make a bunch of spline shapes, but they are all quite simple, sharp edged splines. They never look particularly nice, and they don’t fit the shape of the body all that well. As such, I like to make a lot of my controllers by hand. To do this, I used the mesh itself to guide the shapes. I turned on snap to vertex and created a selection of curves that flowed around the area of the body that I wanted to control. I generally tweak them slightly afterwards to make sure the ends of the curves meet up and dont leave gaps anywhere. With this complete, I had a bunch of individual curves that could be selected seperately. What I actually want is to be able to click anywhere on any of the curves and to have them all selected. To do this, I had to reparent the individual “curveshapes” to a single curve. This can be easily done by selecting the “curveshape”, then shift selecting the curve I want to parent it to. I then simply use a single line of MEL script: “parent -r -s”. This leaves an empty curve node with no shape that can be deleted.

IKSpine02 IKSpine03

Once all my controllers were built I aligned them with the correct bones and parented the clusters to each controller. I also created two groups for each controller to be parented within. One I suffix with _SDK and one with _0. The _0 is my null group. The 0 point so that I do not need to use freeze transformations. The _SDK group allows me to set up parent constraints for a controller, whilst still giving the animator the ability to animate it. For the spine, I parent constrained the middle controller _SDK to both the top and bottom spine controllers. This means that the middle controller will always remain halfway between the top and bottom of the spine.

Once this was complete I decided to test the spine to check it was working correctly. Unfortunately, it wasn’t. I hadn’t realised I had only given the IK spline four bones to move around. When the curve had extreme bends the bones just averaged out their positions and the shape of the curve was lost. This meant rebuilding the deform spine with more bones so that there were enough joints to follow the spline curve more accurately. Having added them, I made sure to tidy up their orientation with comet tools again.

IKSpine04 IKSpine05

I then repeated the process of applying an IK spline to the duplicate set of bones and creating clusters for the three control points of the curve. I re-positioned the controllers to ensure they were correctly aligned with the new bones and then parented the clusters to the controls. I also set the twist controls for the IK spline to make the hip and chest controllers control

the spine rotation.

 

IKSpine07

Finally, I wanted my spine to be stretchy as this is meant to be a “cartoony” rig. I created a multiplydivide node which I set to divide. I also created an arc length info node for the spline curve. This provided me with the length of the curve at any time. I wired the length into the first input of the divide node and put the length of the curve when all controls were at 0, 0, 0 into the second input of the divide. This means that the output will be the current length of the spine divided by the original length. I then simply wired the output into the scale x (the length)  of all the joints in the spine.

IKSpine06

Success! An easy to use stretchy IK spine.

 

 

Usable Rig (at last)

So yesterday, I finally got the elephant rig to a point where it could be referenced into the animator’s files. One of the major things I worked on was updating all of the controllers so that they are clearer and fit the model nicely. I foolishly assumed this would be easy, but I hadn’t reckoned on the awkwardness of the curve creation tools in Maya. It took me quite a while of just repeatedly trying to create shapes and deleting them as they failed to work. I think my determination to create things that were perfectly symmetrical possibly did not help the situation, but an assymetrical controller just doesn’t look as neat and clean in my opinion. Eventually I hit upon the idea of using the snap to vertex tool and using the edges and vertices of the elephant to help me create controllers that fit nicely to the contours of the elephants body. Having drawn a selection of curves I needed to then join all the individual curves together into a single item. This involved reparenting the individual curve shapes a single curve node and then deleting the rest of the empty nodes. Frustratingly I could find no way to tell Maya to actually combine all the shapes nodes on each curve into one single curve, but each controller selects the entire item wherever you click it, so it still works, its just not as clean as I would like it to be. I then scaled the controllers out from the body slightly and coloured them. I had hoped I could then parent these new shapes to the controllers already in existence (as I had with each individual curve to make the new controller), but every time I tried, the new controllers were rotated strangely and moved away from the body. This was due to the difference in positions of the pivots of the old and new controllers. Hoping I could avoid having to reposition each new controller I decided to instead break all the constraints and set everything back up on the new controllers. It turns out I still had to reposition the pivots, and so rearrange the shapes, but at least I knew I didnt have to spend time trying to delete the shapes of the old controllers, I could just remove the entire item.

I did, however, forget to redirect the spine rotation to the new controllers, so I had a bit of a scare later in the evening when I created a global control and tried to check that everything moved as I wanted it to. When the elephant rotated 90 degrees, the spine flipped, presenting a problem I had first encountered in my 2nd year when rigging a quadroped in 3ds Max. I panicked for a while that my IK spline spine was in fact broken and I would have to come up with a completely new set up. However after I checked the IK I realised that in creating the new controllers, I had not told it to use them to decide the rotation of the spine. Thankfully, this fixed the problem.

IKspine03a

I also needed to update the rig with the new low poly model that Paul had altered for me. I brought the mesh in and whilst trying to work out how to load the skinning from the old mesh to the new mesh, I found an option that instead replaced an old mesh with a new mesh. I tried it out and it worked brilliantly. The old mesh changed to the new mesh. However, I now had two versions of the new mesh, one that was skinned, and one that was not. Assuming that the unskinned mesh was no longer needed I promptly deleted it. A couple of hours later, when testing some other part of the rig, I discovered my mesh no longer seemed to be moving with the bones. Confused I saved the file under a new name, closed it and reopened it. To my horror, the mesh was now invisible. The outliner still showed all the various parts of the mesh, but I couldn’t get them to appear.

MissingMesha

I hastily opened my previous iteration only to discover that that file suddenly had exactly the same problem. Desperately hoping I hadn’t somehow broken every single version (and so lost all my skinning) I tried the next step back. To my relief the old mesh was there and skinned and working absolutely fine. I had simply lost my day’s rigging work, but nothing else. Deciding that replacing the mesh clearly wasn’t the best method to update my rig, I started working on saving off the skinning so that I could load it onto the new mesh. Frustratingly it seemed Maya was only giving me the option to load each bones skinning one at a time. It was doable, but a bit pointlessly time consuming. Fortunately, I knew one of my classmates, Joe, had successfully, and easily, loaded skinning onto new meshes during his project. I asked him about it and he showed me a quick and easy method. It involved skinning the new mesh to the bones (but not editing it at all) and then telling the new mesh to look at the old mesh for the skinning values. Maya can load the skinning in a variety of ways, by volume, by UV map etc. It was brilliant and loaded the skinning onto the new mesh perfectly. I didn’t even need to tweak it, though Joe had warned me I might need to. This is great to know as I now know I can quickly skin the high poly elephant to the rig (and tidy it up afterwards) as soon as it is ready. I will not have to go through the time consuming process of skinning from scratch again.

The last thing I needed to build was dynamic tail. Having already gone through the long process of working out how to do the trunk, it was simple a case of repeating the method on a much simpler chain. The dynamic output curve became a blendshape for the spline whilst the controls affected the dynamic input curve. Again, unfortunately, the rig doesn’t update its position until the animation is played, but, to my current understanding of dynamics, there is no way around this.

I also created a control for the tail that will rotate all three FK controllers at the same time. I actually created three of these for the trunk as well, so that an animator can control the entire tail (or a section of trunk) without having to select a whole bunch of controllers. Every controller I create is parented to a group (with the suffix _SDK) and that group is then parented to another group (with the suffix _0). The _0 group becomes the null group, which provides a 0 point for position and rotation. The _SDK group allows me to create batch controllers whilst still making the individual controllers able to tweak the bones position. I simply wired the rotation of the batch controller to the _SDK groups of the relevant individual controllers. When the batch controller is rotated, each _SDK group wired to it also rotates. The individual controllers parented to the _SDK groups also rotate (due to the parenting) and so rotate the relevant bones. However, because the controls are not wired to anything, the animator is still able to tweak the position of the bones individually at any time.

 Tail01a

I then set up a switch for the tail to allow the animator to blend between dynamic and FK. Like the trunk I also set up some attributes to allow the animator to change the stiffness and flexibility of the curve dynamics if they wish.

Tail02a

Finally, I added some empty attributes to various controllers ready to be wired up to blendshapes when I have the highpoly mesh. I created them in advance so that it is less likely there will be any problems with the referencing when I update the rig later on. I wanted to make sure that everything that might be animated was already in place, and so it is only skinning and wiring and not controllers that will change in future files.

The Elephants Trunk

The elephant’s trunk was one of the most challenging parts of the rigging. I knew I wanted a simple FK trunk, and I knew I wanted a trunk with dynamics. However one of my animators had also informed me they wanted an IK spline trunk to animate with as well. As such, I needed a trunk that could switch between any of these, and more importantly, blend between any of them to any of the others. My desire to allow the trunks to blend meant I couldn’t simply use an “enum” attribute to change the parent constraints as this is an “all or nothing” attribute. Theres no way to have half of one and half of another. I also decided I couldn’t use a single “float” attribute with one of each of the control types at -1, 0 and 1. There would be no way for me to blend between the two trunk control types at either end (-1 and 1). They would only be able to each blend with the control type that was at 0.

Eventually I decided to use two different sliders, one that would blend between FK and “other” and a second that would blend between IK and Dynamic. This second slider controlled what the “other” was. In order to do this I had to create four duplicates of the trunk bone set up. I had an FK duplicate, an IK duplicate, a Dynamic duplicate and the “other” duplicate. The deform trunk was parent constrained to both the FK and the “other”. I then wired up the FK/other attribute to control which parent constraint was in use. The “other” trunk was then parent constrained to both the IK and the Dynamic and these parent constrains were wired to the second IK/Dynamic attribute.

It meant spending a huge amount of time in the Hypershade, wiring up various items. It was extremely time consuming as I had to organise items in the hypershade so I could see what was parented to what and where the wires actually needed to go. Its definitely the largest wiring system I’ve created when rigging so far.

Trunk01a

Trunk02a

Trunk03a

Having achieved the ability to blend between any of the different control systems, I needed to actually work out how to create a dynamic trunk. I watched a variety of tutorials and read a load of sites before deciding what I felt would be the cleanest, and simplest, way of rigging it. I applied another IK spline to the trunk and then duplicated the spline curve. This duplicate would be the input for my dynamics. So this was the curve that I had to create controls for. I made them in the same way as I did for the trunk, by creating clusters for the various points along the curve and parent constraining the clusters to controllers. I then applied dynamics to this duplicate curve which created another two curves in the process. One was the dynamic curve, and one was the output of the dynamics. The final step was to make this output curve control the original IK spline. I simply applied a blendshape to the spline, turned it up to 1 and locked the attribute so it couldn’t be changed or broken.

Working with dynamics is slightly strange though, as the dynamics only update when Maya plays the animation. This means you can tweak the starting position of the trunk, but it doesnt actually move the mesh (or the bones) until you press play. I think it unlikely the animators will want to ever just have a dynamic trunk, but I feel it may be nice to blend with some FK animation to add some extra secondary motion to the trunk.

Finally, I created some extra attributes so that the animators can edit the flexibility and stiffness of the dynamics and so affect the way Maya calculates the shape of the output curve.

The Basic Rig

The basic rig is now complete, which means I can now create the deform test and actually skin the low poly mesh to check how everything deforms.

Deform Skeleton

It was such a relief to finally be doing something that felt constructive. My first step was to build the deform skeleton. This is the set of bones that the mesh will be skinned to. It is important to get the bones placed correctly so that joints bend in the right place and the deformations of the mesh are as natural as possible. With the skeleton built, I used comet tools to help get all the bones aligned correctly. This ensures that the axes of rotation are where I want them to be, and therefore the rig will be easier/neater to use.

At this point I also made sure all my bones were named correctly. Every joint name starts with either C_, L_ or R_. This denotes whether the bone is down the centre of the body (C_), on the left hand side (L_) or on the right hand side (R_). This is then followed by the name of the bone, and possibly a number (ie spine01 or trunk03). Finally, every joint ends with _jnt. A good naming convention is essential when rigging as it makes things much easier to find with a search tool etc.

IK Spline Spine

Once everything was named correctly I set to work creating a control system for the spine. I used an IK spline set up. To avoid causing problems and having to rebuild the deform skeleton, I duplicated the spine and placed the IK on the duplicate. I could then parent constrain each individual IK bone to its equivalent on the deform skeleton. This causes the deform skeleton spine to do whatever my IK spine does. I chose to create an IK spline with just two spans. This means the spline has three control points: one at either end and one in the centre. I turned these into three clusters which I could then parent to various controls.

IKspine01a

I created a simple cube control for the hips and shoulders of the elephant which I deformed slightly so that they fit the contours of the elephant slightly better. I then parent constrained the corresponding clusters to these controls. I used a simple circle for the middle of the spine and again parent constrained the centre cluster to this control. This provided a way to control the spline curve and so control the bones. However, if a control was moved too far away and the spline curve became too long, the bones did not stretch and so only covered part of the curve.

This was an undesirable result, so I set up a fairly simple wiring system in the hypershade. I created a node for the spline curve with a single attribute: the arclength of the curve. This value changed whenever the curve changed, meaning I could always know how long the curve was. Using this changing value, and the curve’s original length, I set up a divide node that would calculate the current length divided by the original length. Then I used a condition node that was connected to the x-scale (length) of all the bones in the spine. If the curve was the same length or shorter then the scale of the bones remained at 1. However if the length of the curve was ever longer than its original length, the scale (length) of the bones was changed from 1 to whatever the divide node calculated. This meant if the controls were ever moved too far apart, the bones would scale and still fill the entire curve.

IKspine02a

However, I discovered that while this worked really well when the curve was long, it didn’t work very well if the curve got too short. Some of the bones started flipping to try and remain within the length of the curve. As such I decided to completely remove the condition node and simply wired the divide node straight into the x-scale of the bones. The spine now stretches and contracts brilliantly.

IK leg

The next step was to create an IK leg set up so that the legs could be easily animated. I debated for a while which two bones to put the IK on before deciding it definitely needed to be shoulder->wrist/hip->ankle. Like the spine, I duplicated the deform bones and applied the IK to the duplicates. I created a foot controller which was, again, a slightly deformed cube. I parented the IK target to this controller and orient constrained the ankle joint. This meant that I could move the controller and the leg would bend, but the foot would remain in whatever orientation the controller was in. Finally I added a controller to move the top of the leg around. I then repeated this for all four legs.

BacklegIK01a

The rest of the body I decided to keep simple for this basic rig and I used simple FK set ups in the neck, head, trunk, ears and tail. With all the controls complete I decided to try and make the rig clearer to use. I started colour coding my controls; green for left, red for right, cyan for anything down the centre. I also made a centre of gravity controller which could move both the hips and shoulders at the same time, in case the animators wanted some FK control from the hips. This I coloured yellow to make it stand out.

BasicRiga

However I decided that the cyan was too close in colour to the “selected” colouring that Maya uses. As such I changed things again:

BasicRig02a