Editing Motion Capture Data

This chapter describes several techniques for modifying motion capture data, including methods for layering artist specified motion, decomposing motion capture data into rotational components or poses, and blending poses into motion capture. It also describes how to use the Motion Transform command to use the same motion data on skeletons of different sizes and proportions.

---

In this Chapter

• layer artist's motion with motion capture data
• use the Motion Transform operation, which lets you use motion capture data from one skeleton to animate a second skeleton
• decompose motion data into rotational or pose data
• write out decomposed motion data into an .amc file
• use "Difference Posing" to quickly alter saved motion capture data
• use the N-Dynamics Lock Feet operation to lock your character's feet at or below an imaginary ground plane

---

Layering Artist's Motion with
Motion Capture Data

Suppose, for example, that you were animating a character using a motion capture file for a "run." You may have gone to a motion capture facility, captured a good run cycle, and applied it to your skeleton; however, you decide later that it would have been nice if you could make the character wave his arms, turn his head, or reach out for an object as he passes by. What do you do then?

Rather than recapturing an entire motion sequence, you could layer artist created motions using the Skeletal Animation System to add these "secondary" motions to the captured motion data.

---

Layering Rotations with
Motion Capture

Try this:

1. Load the following script:

/usr/local/ngc/demo/scripts/layer-mc-and-rotations

In this script, we'll use motion capture data from Acclaim to animate the skeleton.

This script has four channels:

• A Set View channel, which positions the camera.

  
  

• A Read Acclaim Data channel which reads the motion capture used to animate the skeleton.

  
  

• A user-named parent channel, called "Layered Bone Rotations," which contains several children channels of bone rotations.

  Note that this channel is initially deactivated; we'll turn it on later to show you the effect of the layered motion.

  
  
• An Update Skeleton channel, which applies both the motion capture data and the layered bone rotations to the skeleton.

Figure 10.1 A script layering artist motion with motion capture data

2. (CLICK-L) on the Animate button in the N-Dynamics window.

The skeleton walks forward toward the camera, with no extra movements of any kind.

This makes the channel active. Note that the text changes from black to white (text color indicates the active status of the channel).

Note that this time, the skeleton lifts its arms as it walks; this motion was layered on top of the motion capture data read in by the Read Acclaim Data channel.

5. (CLICK-L) on the Layered Bone Rotations channel.

This opens the parent channel so that you can see the subchannels it contains.

In this script, we created channels that added rotation to both the left upper arm and the right upper arm.

The black triangle in the upper right corner of the channel tells you that the channel contains addition subchannels which are not currently visible.

The two vertical lines in the channel are "cues," which are described in more detail in the N-Dynamics Reference Guide. They represent a value for the rotation operation at a specific point in time.

The curve data looks like this:

8. (CLICK-M) on the cue to see its value.

A dialog box shows the current value for the selected cue (at frame 16):

As described in the Skeletal Animation System Reference Guide, all the modifications made to a skeleton (including applied motion capture and rotations) are additive. This means at frame 16, 112 degrees will be added to the Z rotation of the left upper arm.

Try this:

Move the cursor over the channel and press "a".

The skeleton should look something like this:

11. Reactivate the Layered Bone Rotations channel.

Move the cursor over the channel and press "a".

The skeleton should now have his left arm raised:

Things to Remember about Layering Motion

• Layered bone rotations can be applied to any bone on the skeleton.
• Rotations should be applied before the Update Skeleton operation.
• Layered rotations are additive; if you have two rotation channels for the same bone, the two are added together.
• It may be handy (but not necessary) to nest rotations and other user-defined modifications in one channel. In this sample script, we created a parent channel with an operation class of None, then assigned it a Name of Layered Bone Rotations. For more information on building scripts, see the N-Dynamics Reference Guide.

---

Layering IK Moves with
Motion Capture

In this script, we make the skeleton reach out for a panel as it walks by. You could use a script like this, for example, to make a character running down a hallway reach out for some object as it walks past.

Try this:

1. Load the following script:

/usr/local/ngc/demo/scripts/layer-mc-and-IK

In this script, we'll use motion capture data from Acclaim to animate the skeleton.

This script has four channels:

• A Set View channel, which positions the camera.

  
  

• A Read Acclaim Data channel which reads the motion capture used to animate the skeleton.

  
  

• An Update Skeleton channel, which applies the motion capture data to the skeleton.

  
  

• An IK Move to Object channel, which makes a specified bone on a skeleton (like a hand or foot) reach for an object

  Note that this channel is initially deactivated; we'll turn it on later to show you the effect of the layered motion.

  
  
• A Make Visible channel to make the object the skeleton is reaching for visible

Figure 10.8 A script layering an IK move with motion capture data

2. (CLICK-L) on the Animate button in the N-Dynamics window.

You'll see the skeleton and the small panel in the N-Geometry window, but as the skeleton goes by, he does nothing. We want the skeleton to reach out for the panel.

This makes the channel active. Note that the text changes from black to white (text color indicates the active status of the channel).

Note that this time, the skeleton reaches out for the panel; this IK move was layered on top of the motion capture data read in by the Read Acclaim Data channel.

Note. The IK Move to Object operation should come after the Update Skeleton operation.

5. (CLICK-M) on the IK Move to Object channel.

In the channel, you specify the bone to move (in this case, the left hand), the DOF set used to complete the IK move, and the target object you are reaching for.

Close the channel when you're done.

We used a linear curve with two cues:

if you (CLICK-M) on the first cue, you can see that its value is 1.0.

This cue is located at frame 33. This means that at frame 33, the specified bone will have moved all the way to the specified object in the script.

The second cue in the script (the one at frame 76) is set to a value of 0. From this point until the end of the script, the IK Move to Object operation has no more effect (since the operation has a value of 0.

As noted before, remember that any IK operations should come after the Update Skeleton operation in your script.

---

Copying Motion Data between Skeletons

Note. Because motion capture data is bone-name dependent, you should only use this operation to copy motion data between skeletons from the same vendor.

1. Read in the skeleton for the motion capture file you want to animate.

Figure 10.12 Read in the skeleton from the motion capture file you want to animate

2. Read in the skin you want to animate.

3. Align the hips of the skin with the root of the skeleton.

Figure 10.13 Move the skin so that its hips are aligned with the skeleton's root

4. (SHIFT-L) on the skin, then (CLICK-L) an Make Displacement.

Make an Absolute displacement on the skin and call it "Base". (This is described in the N-Geometry Reference Guide.)

Note. Moving the root of the skeleton out of the global origin will likely cause unwanted behavior when interpolating the motion data. If it is necessary to move the skeleton away from global zero, you should do so in a dynamics transform-root channel.)

5. Using N-Geometry's skeletal operations, modify the skeleton so that it fits your skin.

Use the various rotate and scaling commands, or the pose editor, to match the skeleton to the skin.

6. Save a new base state for the modified skeleton.

(SHIFT-L) on the skeleton, then (CLICK-L) on Base. Use the same name for the base state of the skeleton that you used for the skin in step 4.

(CLICK-L) on the Done button to save the base state.

7. Rename your altered skeleton with a distinct name.

• (CLICK-L) on GeoMenus>Rename.

  
  

• (CLICK-L) on the skeleton.

  
  

• Enter a new name for the skeleton in the dialog box that appears

  

At this point, you have two elements in memory:

• The skin
• The altered skeleton

To complete the transform, you must reread the original skeleton back into N-Geometry:

8. Reread the original skeleton from the motion capture file.

You should see both skeletons in the N-Geometry window:

9. (SHIFT-L) on the altered skeleton, then (CLICK-L) on Motion Transform.

10. Select the original (unaltered) skeleton from the object list that appears.

You're next asked if you want to select a bone on which to base the transformation of the translation data.

11. (CLICK-L) on Yes.

12. You're next prompted to select the bone on the altered skeleton that the translations should be based on.

• For a walk cycle, select a femur.

  
  

• For motion data in which the arms generate the movement (e.g., swinging from vine to vine), select an upper arm.

Note. When doing a Motion Transform, try to keep the bones which are dependent on for the motion transform in close proportion to the same bones on the original skeleton. You can still scale the entire skeleton (e.g., to copy the motion from a man to a dwarf), but don't change the proportion of the bone to the overall skeleton.

For example, if you are doing a walk cycle and do not want any slippage or Y translation, try to keep the bones of the legs in the altered skeleton proportionate to the bones of the legs in the original skeleton. If you need more consistency in the motion of the arms, then keep the upper arm bones proportionate.

In general, avoid using any Scaling operations on just the bone being used on the motion transform.

Once you've completed the transform:

13. Create an N-Dynamics script with at least two channels:

• Read Acclaim Data or Read Biovision Data

  
  

• Update Skeleton

If you're working with an Acclaim skeleton, use Read Acclaim Data, and complete it like this:

• For Skeleton, specify the altered skeleton

  
  

• For the .asf file, specify the original skeleton.

  
  

• For the .amc file, choose the motion file.

If you're working with a Biovision skeleton, use Read Biovision Data, and complete it like this:

• For Skeleton, specify the altered skeleton

  
  

• For the .bvh file, choose the motion file.

In the Update Skeleton channel, make sure you select the altered skeleton:

16. (CLICK-L) on the Animate button to animate your script.

After testing the motion transform on the new skeleton, you can proceed to skin the skeleton as described in the following chapters.

Note. Moving the root of the skeleton out of the global origin will likely cause unwanted behavior when interpolating the motion data. If it is necessary to move the skeleton away from global zero, you should do so in a dynamics transform-root channel.

---

Decomposing Motion Data

• rotational components (for each bone on the skeleton)
• individual poses (as in the Animate Pose to Pose operation)

Typically, this operation is performed on a channel containing one of the following skeletal operations:

• Read Acclaim Data
• Read Biovision Data
• Read Motion Analysis Data
• Animate Pose to Pose

As described later on in this section, animation data that has been decomposed into poses can be further decomposed into their rotational components.

Decomposing Motion Data into
Rotational Components

1. Add a channel to your script that reads motion capture data for a skeleton.

2. Specify the skeleton and any motion capture file you want to decompose.

For an Acclaim skeleton, specify both the .asf and .amc files; for a BioVision skeleton, specify the .bvh file.

The Action Operations menu appears.

The following menu is displayed:

5. (CLICK-L) on Separate rotation channels.

When you choose this option several things occur:

• The channel changes from a Read Acclaim Data or Read Biovision Data into a Decomposed Motion channel (where Motion is the name of the .bvh or .amc file).

  
  

• Subchannels are created for different parts of the skeleton: arms, legs, other bones, and a separate channel for the root transforms.

Open the script to see the rotational components for each bone:

6. Open any of the channels to see the bones included in that channel.

Figure 10.24 Motion capture file decomposed-showing separate bones

You can continue opening channels until the actual curve data for a given bone is displayed; this curve data can be manipulated just like any other curve data in N-Dynamics:

Decomposing Motion Data into Poses

1. Add a channel to your script that reads motion capture data for a skeleton.

2. Specify the skeleton and any motion capture file you want to decompose.

For an Acclaim skeleton, specify both the .asf and .amc files; for a BioVision skeleton, specify the .bvh file.

The Action Operations menu appears.

The following menu is displayed:

5. (CLICK-L) on Pose to pose.

The following dialog box appears:

6. Enter the number of frames between each pose.

When you choose this option, several things occur:

• The channel changes from a Read Acclaim Data or Read Biovision Data into a Decomposed motion channel (where motion is the name of the .bvh or .amc file).

  
  

• A single pose to pose subchannels is inserted into that channel. Cues are inserted evenly spaced over the length of the channel.

Note. The values shown for the cues are "dummy" values (as described in the Animate Pose to Pose operation). The peaks and valleys are created only to make it easier to see each cue.

Open the script to see the cues that are inserted for each pose:

Figure 10.28 Motion capture file decomposed into poses

Decompose Motion Data from Pose to Pose into Rotational Data

1. (CLICK-R) on the Decomposed motion channel.

2. (CLICK-L) on the Decompose command.

3. (CLICK-L) on Separate rotational channels.

Doing this converts the data in the pose to pose channel into rotational component subchannels, much like those described in Figure 10.23; however, note that the cues inserted into the subchannels are preserved, and that a marker track is inserted for the channel, which makes it easy to slide poses in time and still be able to edit their values:

Sliding the individual cues slightly left or right in such a decomposed channel can lessen the "robotic" feel of some motion capture files.

---

Writing out Your Own Motion Data

Then, you can make a simpler script that uses only the generated motion capture data to drive the skeleton, using a simple Read Acclaim Data and Update Skeleton combination.

This operation might be used, for example, after you had used the IK Move along Wire operation to tweak a previously saved motion capture (.amc) file, or after you've used the difference posing technique described elsewhere in this chapter.

To use this operation:

1. Insert a Write Acclaim Data channel at the bottom of your script.

Note. This operation should come after the Update Skeleton operation.

2. (CLICK-M) on the channel.

In the dialog box that appears, specify the skeleton from which the .amc file is to be generated and where the file is to be saved. The parameters in this dialog box are described in the table below:

Table 10.1 Write Acclaim Data parameters

Parameter	Dynamic?	Description
Skeleton	No	The skeleton being used to generate the Acclaim data.
Skeleton Dir.	No	The .asf file with which the .amc file should be associated.
.asf File Name	No	The name of the .asf file.
Motion Dir.	No	The directory into which the motion capture file should be written.
.amc File Name	No	The name of the motion capture file.

The .asf and .amc files are written out.

Note. If you write out data from a frame range in the middle of the script, written frames are numbered 1 through N, regardless of their frame number in the script.

This lets you write out portions of an animation to a new .amc file.

---

Difference Posing

You could use this option, for example, if you were animating a skeleton with motion capture data for a simple walk cycle, then decided you wanted the skeleton to lean forward under a door. You first animate the walk cycle to determine where the skeleton should "duck," then pose the skeleton to its desired position, then blend that difference pose back into the script.

To create and use a difference pose, you'll usually start with a script that contains two channels:

• One channel with a Read Acclaim Data or Read BioVision Data operation.
• Another channel with an Update Skeleton channel, which applies the motion capture data to the skeleton.

A sample script might look something like this:

Figure 10.30 Sample script

(CLICK-L) on the Animate button in N-Dynamics to animate the script. This will give you an idea of where you want to make the change. Suppose, for example, that your skeleton is upright at frame 50:

Figure 10.31 Skeleton upright at desired frame

Your animation calls for him to duck under a doorway at frame 50; with difference posing, you can make this change very quickly. You need to use one of three operations to use difference posing:

• Animate Pose
• Animate Multiple Poses
• Animate Pose to Pose

All three of these operations let you copy poses into cues; the first two are similar, and use difference posing in the same way; the third is slightly different and is described in a different section below.

Difference Posing with Animate Pose (or Animate Multiple Poses)

1. (CLICK-M) on the Update Skeleton channel and make sure that the Save-Pose option is set to Yes.

Figure 10.32 Set Save-Pose to Yes

When Save-Pose is set to Yes, the skeleton's pose is temporarily saved at each frame. This temporary pose can then be referenced when creating a difference pose, and can be saved with the skeleton like any other pose.

This creates a "temporary" pose that you'll base the difference pose on.

4. Reorder the channels so that the new channel comes before the Update Skeleton operation.

The channel can go anywhere before the Update Skeleton channel; however, it's probably most handy to put it under the channel whose motion you are trying to modify.

6. Specify the skeleton you want to pose.

7. (CLICK-L) on Pose.

The following menu appears:

At this point, your skeleton should be animated at the frame where you want to modify it. Don't click yet!

Suppose, for example, we want him to duck beneath the doorway:

9. Now, (CLICK-L) on Copy and Save Difference Pose.

The following dialog box appears:

10. (CLICK-L) on Last Animated Pose.

This lets you specify which pose this "difference pose" should be measured from.

Figure 10.36 Specify a name for the difference pose or use the default

(CLICK-L) on Done after specifying the name, then (CLICK-L) on Do it to save the pose.

Lets assume it's at frame 50; your script should look like this:

13. (CLICK-M) on the cue to edit its value.

Specify a value of 1.0 if you want the skeleton to fully execute the pose.

You can either (CLICK-L) on the value (to bring up the keypad) or use the slider to change the value of the cue.

The difference pose is now incorporated into the animation.

15. (CLICK-M) on the curve channel under the Animate Pose channel, then (CLICK-L) on the Curve type text edit box.

A list of curve types appears:

Choosing one of the other curve types changes how the pose is layered over the motion capture data; for example, a Slow-In/Out curve makes the transition more gradual than the default Linear. Try several different curve types, reanimating after each change, until you get the results you want.

The Animate Pose to Pose operation lets you specify several poses which can be layered over motion capture data.

Figure 10.40 Set Save-Pose to Yes

When Save-Pose is set to Yes, a "temporary" pose is saved for the specified skeleton each time you animate a frame. This temporary pose can then be referenced when creating a difference pose. (Unlike the Animate Pose operation, this temporary pose is not saved.)

This creates a "temporary" pose that you'll base the difference pose on.

4. Reorder the channels so that the new channel is after the channel that reads in themotion capture data.

The channel can actually go anywhere before the Update Skeleton channel; however, it's probably most handy to put it under the channel whose motion you are trying to modify.

6. (CLICK-L) on Skeleton and specify the skeleton you want to pose.

7. (CLICK-L) on bones, make sure all the bones are selected, then (CLICK-L) on Do it.

At this point, your skeleton should be animated at the frame where you want to modify it.

Lets assume it's at frame 50; your script should look like this:

9. (CLICK-M) on the cue to edit its value.

The following menu appears:

10. In the N-Geometry window, pose the skeleton as he should appear at that frame.

Suppose, for example, we want him to duck beneath the doorway:

11. Now, (CLICK-L) on Copy Difference Pose.

The following dialog box appears:

12. (CLICK-L) on Last Animated Pose.

In the dialog box that appears, the default pose name temp appears.

14. (CLICK-L) on the Animate button in N-Dynamics to reanimate the script.

The difference pose is now incorporated into the animation.

Copyright © 1996, Nichimen Graphics Corporation. All rights reserved.