In addition to being balanced physically,your animation also needs to be balanced in time. If an action occurs—say, a character jumping across a small gap—that action needs to balanced in time both before and after it happens.In a case like this,it’s easy to visualize what we need to do. Before the character jumps, it will coil like a spring, drawing its arms behind it, then swinging them forward to build momentum.The character then flies through the air. After its feet touch down on the other side, its legs will bend to absorb the shock,maybe even taking a step or two forward. The arms will continue forward after the body stops,probably swinging up. The motion before the jump is anticipation. After the jump is the follow through.
Let’s step through the poses in our whole animation project and determine if, when,and how to add these elements.For now,we’ll ignore the walking poses as they are a special case.
We’re looking for changes of momentum: stops, starts, and rapid changes in direction. The larger the change the more anticipation and follow- through we’ll have to add.
I find the following spots where Junot’s momentum changes:
The transition from walking to regarding the pedestal
When she begins to fall backwards after the alarm
The transition from falling backwards to the defensive crouch
Note that each of these points in the animation is a time of change in motion. She is transitioning from one kind of motion to another. We are looking for beginnings and ending.
One other place deserves a bit of anticipation too: Junot touching the light. It isn’t a full body motion, but it is a deliberate, distinct action on her part and adding anticipation here is directorial. It assists the viewer. A proper anticipatory motion prefocuses the viewer on the hand and arm, giving them a cue that “Hey-something’s going to happen here! Pay attention!” It doesn’t need to be huge- remember that an anticipation should be in proportion to the eventual action- maybe only a few frames,but including it is important.
The transition to the backwards fall is going to be trickier. It is not a deliberate motion. Junot did not “decide” to start going backwards. She was scared into it. Think about time when you’ve walked around a corner and come literally face to face with someone. Generally, you jump or twitch (or in my case have a mini freakout) Depending on how good your memory is of what your body does, you might recall that your nervous systems sends a kind of shock into your muscles that causes them to tense at once, propelling you away from “danger.” That is what’s happening to Junot. Strictly speaking then, this motion shouldn’t have any anticipation. It’s an explosion of movement. We could animate it that way, as though the alarm delivered a shock wave that blasted her backward. That would be a valid approach and it’s probably worth a try.
As we’re just learning though, we’re going to choose a method and stick to it. If you want to go back and try it differently later, then by all means proceed. What we’ll do is provide her with only a frame or two of frightened anticipation. If she had a face, this is where we’d stick a wild-eyed response (in animation terms that’s called a “take”) Since it’s going to be a brief, we’ll have to make the pose rather extreme.
When she finally falls into the defensive crouch, it is the end of the careening motion. We’ll need some serious follow through that ends with her hands in the martial arts pose. Follow through poses have always been a little easier to generate for me because they are often simple physics. How would the masses involved need to move in order to return the character to balance?
Let’s go back and work on the first example. Fig 5.3 shows the last walking pose and the standing pose respectively. First, do we need to add anticipation or follow-through? The energy shifts of stopping action (walking to standing still), so we need the latter. Junot is moving at a fairly slow place-normal walking- but it is her whole body makes the change. In real life we ease into a stand still. But this is the world of animation and we have to liven things up.
A follow-through that we need here will actually occur between our final pose (standing,regarding) and the earlier one.The resting pose is the one that occurs after the follow-through,not before it. So really, when I was story boarding, I missed two key motions. First Junot walks. Then she begins to stop. There is a follow through pose.Finally,she comes to rest in the “regarding” pose that actually made it into the storyboards. For this exercise,let’s start with another duplicate of our animation file.The timed constant interpolation steps are pretty good, and again,we’d like to keep a fall back point on hand,in case we completely mess things up.The web site has a file called chapter 05scene.blend to work with as well. I begin to work on the Animation screen. Junot, her control armature, and ground are shown in local mode in the 3D view (\-key), and the Dope Sheet is set to Action Editor mode.We’re still using constant interpolation. It’s often easiest to build a follow- through pose from the final resting pose. To do this,select any of the keys in the column that represents the “at rest” pose. In the sample file,thesearethekeysonFrame61.Press Alt-RMB on any oftheFrame61keystoselectallofthekeysonthatframe.Then, press Shift-D to duplicate the keys and move them four frames or so to the left: backwards in time. Click the LMB to accept the duplication and transformation. Fig. 5.4 shows the Action Editor with the column of keys duplicated and offset. In the figure, the selected column is the new one. It is on this new frame (Frame57) that we build the follow-through pose. Remember to exaggerate and use solid drawing techniques.Keep the character in balance.However,this is animation ,and we can afford to throw in a little bit of the goofiness.I’m going to try to have her feet stop in place, but have the upper body continue forward just a touch so that when she falls into her final pose it’ll be a “Whoa there!” effect, the upper body rubber- banding back into place. Before building the pose,make sure that the current frame marker is on the duplicated, earlier set of keys on Frame 57. Fig. 5.5 highlights a control on the Timeline view’s header.This “record” style button turns on Automatic key framing. We’ve been setting keys by hand up to this point, but now it becomes useful to have Blender take care of that for us.
We already have an entire column of keys, and it would be handy to just have every tweak or change we make on this frame to simply be recorded without any further intervention. So, enable the button. Once you do, ensure that you stay aware of it. Any changes you make to anything in the 3D view will now be recorded in time. This means that if you adjust the angle of the camera,it will record it as a key,not as the absolute,all-time location and rotation of the camera.This can be annoying however, like when you adjust a camera position and find that it is now animated because it has two different positions in time.To avoid this annoyance,open the User Preferences, head to the Editing tab and enable the Only Insert Available option. This will only place keys for objects (or bones) that already have keys, even during automatic key framing, meaning that only things that you have hand-applied a key to(with the I-key)will receive automatic key frames. Something like a camera that has never been keyed will not receive them automatically even if you adjust its position. The first thing I do when actually creating the pose is to go into a top view and find out the direction of the character.By toggling back and forth between the walk and stand,the main direction of the body becomes obvious. Fig. 5.6 shows both poses,with a line drawn to indicate the overall motion.The upper body controller bone should continue along this line as it should follow the body’s momentum. After that, it’s a matter of thinking about what might happen if you try to stop that momentum. Fig. 5.7 shows a more cartoonish approach to this. The pose is inserted before the resting pose because the character hits the position and stops,does this bit of follow-through, then returns to the rest pose. It’s a nice pose,but in the end I’m not sure how well it fits with the action and character in the rest of the animation.It’s kind of goofball. To try a different pose,I use the Alt-RMB to select the entire column of keys that represent the pose and move it way off to one side,out of the normal timeline for this animation.This way,I can easily get it back if I change my mind.Realizing that our rest pose is actually the endpoint of the follow-through,it might be a good idea to generate an additional pose to help us.This new pose should once again fall between the last walking pose and the resting pose,but this time it will indicate the character’s attempt to stop moving. We can just duplicate the resting poses over all frames to the left again and alter it.In this new pose shown in Fig. 5.8, Junot’s left foot is part of the copied pose, in it’s final spot.
The right foot, however, has not moved from the earlier walking pose. This is because it is really her next “step,” and the planted forward foot does not move. To do this, I selected the foot controller in the 3D view and found that the corresponding animation channel in the Action Editor was highlighted. In this case, it is called heel_R. I select the key from the earlier column(the last walk pose),duplicate it with Shift-D, and move it forward to takeover the key in the current pose’s column. Fig. 5.9 shows the selected key, and where it moves to. Doing this guarantees that the foot will be in the same place as it was in the earlier set of keys. Then,I adjust the foot’s rotation so that it is flat on the floor. Stepping back in time, I see where the next swing of the arms would be and move them accordingly. This isn’t the world’s most exciting pose, but that is okay for once. It is merely transitional.The next pose we create will show the follow-through.The one after that is our existing rest pose, with Junot considering the pedestal.
The same rules apply to this attempt at a follow-through pose as to the last. Follow the momentum of the body by going backwards and forward in time. We keep the upper body controller moving just a little along the main line of action, bend the spine slightly, and pull the arms ahead of the body. The resulting sequence can be seen in Fig 5.10
Put together like this, the action of Junot coming to a stop, her body pushing a little past it (follow-through!), then resolving into her rest pose, is obvious. One of the lessons to take away from this is that you cannot always rely on your main poses as a guide to what you’re doing. The pose that you think might require some follow-through might actually be the end point of the follow-through, like it is here.
Let’s move onto the anticipation right before Junot explodes backwards. Anticipation doesn’t have to get nearly as fancy as follow-through, and you have more freedom. Sometimes you need to really watch the physics. Most times,it is just an anticipatory action that draws attention and prepares the viewer for the main action. Fig.5.11 shows the “before and after” poses.We need to put an anticipation pose right before the “b” pose.This one is pretty simple.We give Junot a very quick,light movement toward the pedestal to compensate for the upcoming backward motion. It will only show for a frame or two,giving a subtle but useful effect. Although we’re showing this pose in the long shot for consistency, be sure to build it from the medium shot camera’s point of view, as that’s where it will be captured in the final render.
I’m going to build the pose only three frames before the first “falling backwards” pose, whichcomesonFrame241.So,I duplicate the earlier pose (during the anticipation she won’t have moved her feet)and bring the new column of keys up to Frame238. Keeping in mind where she’s headed, some minor adjustments are all that are needed.The goal is to provide motion contrary to that of the action that is about to happen. By noting the differences in the two earlier figures,we can see that the right arm moves the most.So,it should also receive the greatest degree of anticipatory movement. Fig. 5.12 shows the anticipatory pose almost involuntary action, so the anticipation will be reduced.
If Junot had been leaning against the pedestal and actually planning such a movement, the anticipation would have been significantly greater. We can’t tell exactly how it’s going to look until we release everything from constant to smooth interpolation, but at this point, animation is like a chess game. We’re trying to get all of our pieces into powerful position, with the proper lines showing. If we do that consistently, we’ll find that we are in a very good place later in the game.
The last major pose we need to add is a follow-through pose in the transition from the backwards fall to the defensive crouch. Much like the earlier situation, it appears that the defensive crouch itself does not require follow-thorough after it, but before it. This is because the crouching pose represents the character once it has regained control of itself, after the initial stop and follow-through. So, we need to add two more poses: a stop, and then some follow-through. Fig 5.13 shows the last falling pose and the crouch, respectively. Like earlier, we find the keys that represent the crouch, duplicate them and move them to the left in the timeline. In this new pose, Junot should be trying to catch her balance. The hips and center of gravity won’t be quite as slow as the resting pose, and the hands are not yet up in the defensive position. This is a transitional pose, so it probably won’t have the dynamism and visual impact of one of our main poses. For now, I place this new pose 12 frames before the resting pose. That will give the entire follow-through process, from beginning to recover, half a second. It might work or not, but we’ll have to wit until we release to smooth interpolation to find out. Fig 5.14 shows this pose. It’s not our follow-through, but represents where Junot is trying to end up. Think of it this way. Our storyboard indicated an action pose and a resting pose. What we need to have in the animation is an action pose, a pose on which the character thinks they are stopping, some follow-through, and then the resting pose. To take up the energy of her backward fall, the upper body control goes way down toward the ground, continuing its current momentum. This energy applies itself to her legs which bend deeply, push against the ground, and attempt to counter it. The spine alse curls forward significant, and the arms fly ahead as well, each trying to balance her shifting center of gravity. Fig. 5.15 shows the point of maximum impact. Note that this is not the follow-through. Finally, in Fig. 5.16, we see the follow-through.It’s the continuation of motion past the extreme pose of the earlier figure. It is from here that Junot will transition into the defensive crouch.Sometimes doing an adequate follow-through or anticipation study necessitates more than a single pose. To review,the main rules for building anticipation and follow-through are as follows:
• Actually remember to include them,
• Use anticipation to direct the viewer,
• Make them proportional to the main action,
• Momentum should lineup with the main action
Roland Hess has been working with graphics and imaging software for over 20 years, Roland is the leading expert for Blender software. As one of a handful of people involved with Blender who is both an active user of the software as well as one of the developers, he brings a unique perspective to Blender instruction that helps to bridge the difficult gap between technical knowledge and artistic endeavor. Hess wrote: Animating with Blender for Focal 2009, and The Essential Blender, No Starch, 2007.