How Motion Capture Works in Modern Video Games

A character in a cutscene shifts her weight, glances away and swallows before she answers, and for a second you forget she is made of polygons. Then the camera cuts to gameplay and the same character jogs like a mannequin on rails. If you have ever wondered why, the answer sits in how motion capture works and, just as much, in where game studios choose not to use it.

I spent a week reading studio talks and asking two animators who work on console games to walk me through their pipeline. What follows is the version I wish someone had given me: what happens on the stage, what happens after, and why the raw capture is only the start.

What motion capture actually records

Motion capture, usually shortened to mocap, records the movement of a real performer and turns it into data that can drive a digital skeleton. It does not record what the performer looks like. It records where key points on their body are, many times per second.

The most common studio method is optical capture. The performer wears a tight suit covered in small reflective markers, typically somewhere between 40 and 60 for the body. Dozens of infrared cameras ring the stage, often 30 to 100 of them, each firing infrared light and filming the reflections at 120 to 240 frames per second. Software triangulates each marker from several camera angles and works out its position in 3D space, usually to within a millimetre or two.

What comes out is a cloud of moving dots. No arms, no legs, just points. Turning that cloud into a character is the real work.

How motion capture works from stage to screen

A typical session follows the same broad steps, whether it is for a sword swing or a ten minute dialogue scene.

Calibration. The crew waves a wand with known markers through the volume so every camera agrees on where everything is. Then the performer stands in a T pose and moves each joint so the system learns their proportions.
Performance. Actors perform on a mostly empty stage with simple props, often wireframe stand-ins for doors, tables or horses. A director watches a rough real-time preview of the characters on a monitor.
Solving. Software fits a skeleton to the marker cloud, frame by frame. This is where swapped or missing markers cause glitches, like a hand that briefly teleports to the opposite hip.
Cleanup. Technical animators fix those glitches, fill gaps where markers were hidden and smooth out jitter. For a busy fight scene, cleanup can take many times longer than the shoot itself.
Retargeting. The cleaned motion is mapped onto the game character, who may be taller, bulkier or not even human. A 170 cm actor might be playing a 2.5 metre troll.
Polish and integration. Animators adjust timing, add exaggeration where it reads better on screen, and wire the result into the game engine.

Faces, fingers and performance capture

Body markers capture big motion well. Faces and fingers are harder because the movements are tiny and markers get hidden constantly.

For faces, most studios now use a head-mounted camera: a small rig on a helmet that films the actor's face from a few centimetres away, sometimes with painted dots on the skin. Software tracks how the face deforms and drives a facial rig with dozens or hundreds of controls. Fingers often get a handful of extra markers or separate gloves, and animators still hand-tune a lot of finger work.

When body, face and voice are captured together in the same take, the industry calls it performance capture. It is why actors in modern story-driven games are credited for performances, not just voices. Ninja Theory famously showed this live at GDC in 2016, driving a Hellblade character in real time from an actor on stage.

Method	How it tracks	Strength	Weakness
Optical markers	Infrared cameras and reflective dots	High precision	Expensive stage, occlusion gaps
Inertial suits	Motion sensors on each limb	Works anywhere, even outdoors	Position drifts over time
Markerless video	AI pose estimation from regular cameras	Cheap, fast setup	Less accurate on fast or hidden motion

Why mocap is not a shortcut

A common assumption among players is that motion capture means the animation is "just recorded," so it must be cheaper and more realistic than hand animation. Neither part holds up well.

On cost, a day on a professional stage with crew, actors and equipment can run into many thousands, and then the cleanup and retargeting bill arrives. On realism, raw capture often looks floatier and less readable than people expect. Real human motion is full of tiny hesitations that look like errors on screen, and real weight does not always feel heavy when the camera is behind a character. Animators routinely push poses further and cut frames out of captured motion so actions read clearly.

That is also why the cutscene and gameplay versions of the same character look so different. In gameplay, responsiveness wins. If a captured dodge takes 18 frames to leave the ground because that is how long a real person takes, the game feels sluggish. So gameplay animations are often shortened, blended or partly hand-keyed. How fast the screen updates plays a part too, as we covered in what frame rate actually does to how a game feels.

Plenty of acclaimed games, especially stylised ones, use little or no mocap at all. Hand-keyed animation lets a character move in ways no human can, and that is a feature, not a limitation.

How to spot motion capture in the games you play

Once you know the signs, you start seeing them everywhere:

Idle fidgets that never quite repeat, like shifting weight or rolling a shoulder, usually come from capture.
Foot sliding during turns is a retargeting or blending artefact, where captured steps do not match the speed the game is moving the character.
Perfectly snappy attacks with exaggerated anticipation tend to be hand-keyed or heavily edited.
Developer leftovers. Unused capture takes sometimes survive in game files, much like the tools described in hidden debug menus and why they survive in finished games.

Crowds and background characters are another tell. Open worlds often reuse a library of a few hundred captured clips across thousands of characters, which is why you might notice two strangers on the same street with identical walks. The same kind of reuse logic drives the systems in how procedural generation builds endless game worlds.

Frequently asked questions

How does motion capture work in games?

A performer wears markers or sensors while cameras or suit sensors record their movement many times per second. Software fits a digital skeleton to that data, animators clean it up, and it is mapped onto the game character. The final animation is usually edited heavily before it ships.

Is motion capture better than hand animation?

Neither is better in every case. Mocap is excellent for subtle, realistic acting in cutscenes. Hand animation is often preferred for responsive gameplay moves and stylised characters, and most big games combine both.

Why do mocap suits have balls on them?

The small balls are reflective markers. Infrared cameras around the stage see them as bright points, and software uses several camera views to calculate where each marker is in 3D space. Their positions over time describe how the performer moved.