Does motion capture always need a special suit?

No. Some systems use markers or wearable sensors, while markerless approaches use images without attached markers.

Can the captured motion be used on any character unchanged?

Usually not. Differences in proportions, joints, and contact relationships can require retargeting and editing.

Is every point in the final file directly observed?

Not necessarily. Some values are inferred by a model, filtered, or filled across missing observations.

AI Glossary / Computer vision and multimodal AI / Beginner

Motion capture

Definition

Motion capture is the recording of a person's or object's movement as digital motion data. A system tracks markers, body landmarks, sensors, or other features over time and converts those observations into positions, orientations, or skeletal motion. The data can drive animation, support movement research, or describe how an object moves through a calibrated space.

What it means

An ordinary video records appearance. Motion capture aims to extract a reusable representation of movement, such as the orientation of an arm or the trajectory of a tool. That representation can be applied to a digital character whose appearance differs from the original performer.

Captured motion is not necessarily a direct recording of every joint. Optical markers sit on a surface, inertial sensors measure quantities related to movement, and markerless systems infer landmarks from images. Software models connect those observations to a body or object representation. Calibration, assumptions, and cleanup therefore influence the final motion data.

How it works

An optical system uses cameras to observe passive reflective or active light-emitting markers, or uses computer vision to detect unmarked landmarks. Multiple calibrated views can reconstruct three-dimensional positions. Inertial systems use wearable sensors and estimation methods to track orientations and movement without requiring every body part to remain visible to cameras.

The workflow establishes a coordinate system, synchronizes observations, calibrates the subject or object, records movement, and solves a suitable model. Missing observations and tracking errors may be corrected or interpolated. For animation, retargeting transfers movement to a different skeleton. For measurement, the processing steps and their uncertainty must remain traceable rather than being hidden by a visually polished result.

Three practical examples

1. A game studio records a performer's walk and adapts the motion to a character with different proportions. Foot contact and balance may need adjustment after retargeting.

2. A product-design team records how a tool moves through a workspace. It uses calibrated object trajectories to compare alternative layouts.

3. A movement-research team records repeated actions with a defined marker protocol. It distinguishes measured marker trajectories from joint quantities calculated by its model.

Think of it this way

Think of transcribing a dance into a sequence of movement instructions rather than preserving only the film of the dancer. Another character can follow those instructions, but its different proportions may change the result. The analogy also has a limit: motion-capture systems do not inherently understand the dancer's intention, and their inferred skeleton is a model rather than the living body itself.

Worked example

A marker is recorded at position (0.20, 0.10, 0.50) meters and, 0.02 seconds later, at (0.23, 0.14, 0.50) meters. Its displacement is (0.03, 0.04, 0) meters. The straight-line distance between observations is 0.05 meters, so the straight-line speed estimate over that interval is 0.05/0.02 = 2.5 meters per second.

This calculation assumes that the positions share a calibrated coordinate system and that the timestamps are correct. It measures the tracked marker's movement, not automatically the exact movement of an underlying anatomical joint.

If the second position was filled in during a tracking gap, the calculated speed partly reflects the interpolation method. If the marker moved relative to the body surface, that motion also enters the result. Keeping these distinctions visible is essential when captured movement is used for analysis rather than only for animation.

Illustrative case study

Fictional case: An animation team captures a short sequence in which a performer picks up a cup. The initial skeleton looks reasonable, but the hand passes through the digital cup when the motion is transferred to a character with longer fingers.

The team checks marker visibility, the hand model, and the relationship between the prop and performer. It then adjusts contact during retargeting and labels the edited animation separately from the original capture. The capture provides valuable timing and movement structure, while the final scene still requires artistic and geometric work. The example does not imply that recorded movement is automatically ready for every target character.

Common misconceptions and mistakes

Motion capture is not synonymous with markerless AI. Marker-based optical and inertial approaches are also widely used, and systems can combine technologies. Conversely, removing markers does not remove calibration, occlusion, or model uncertainty.

A smooth skeleton can contain tracking swaps, filled gaps, or incorrect joint estimates. More cameras or a higher frame rate do not guarantee accuracy without suitable placement, synchronization, and calibration. Do not confuse skin-marker movement with direct bone measurement. For animation, plausible corrections may be acceptable; for quantitative studies, those corrections need to be documented and assessed against the measurement purpose.

Business use

Motion capture can reduce repetitive animation work, create reusable movement libraries, and support engineering analysis of tools or workspaces. Its usefulness depends on the required detail, capture environment, and tolerance for postprocessing.

Specify whether the deliverable is an animated performance, a marker trajectory, a rigid-object pose, or a biomechanical model output. Evaluate the complete workflow, including setup, calibration, cleanup, and export. Test representative occlusions and interactions rather than only easy isolated movements. For recorded people, establish the intended use and handling of the resulting movement data as part of the production process.

Check your understanding

A wrist marker disappears for several frames and software fills the gap. Are the filled positions equivalent to direct observations?

Answer: No. They are estimates based on the gap-filling method and surrounding information, and should be identified accordingly when they affect analysis.

Related terms

Motion analysis
Inverse kinematics
pose estimation
marker tracking
skeletal animation
retargeting
inertial measurement unit
camera calibration

Learn next

Inspect a short motion-capture sequence alongside synchronized video. Identify measured points, solved skeleton motion, and any filled gaps. Then retarget the motion to a character with different proportions and note which contacts require correction. This separates recording quality from the later adaptation process.

Sources

Primary sources: Vicon, “What Is Motion Capture,” https://www.vicon.com/about-us/what-is-motion-capture/ ; Carnegie Mellon Graphics Lab, motion-capture database information, https://mocap.cs.cmu.edu/info.php . Relevant technology and calibration descriptions were reviewed. Vendor performance claims are not adopted as universal guarantees; the marker calculation and animation case are original examples.

Previous: Motion blur Next: Motion coding