Key points
- Motion capture turns movement into a location record for a handful of predefined points.
- What's left isn't video — it's a table of numbers. No clothing, no face, no background sits in that table.
- The table sits apart from the original body, so it can be carried over onto a completely different character. A step that adapts it to the new body's proportions is still needed.
- How many points, and how many lines get written per second, decide how fine the movement comes out.
- These days a table can get built from nothing but webcam footage, no special suit or studio required.
The analogy
A train timetable lists, line after line, which station a train is at and when. That sheet of paper says nothing about what color the train is, how many cars it has, or who's riding inside. Just locations and times.
What motion capture produces is the same kind of sheet. It fixes ahead of time which spots on the body to watch — a shoulder, an elbow, a wrist — and at very short intervals, writes down one line for where those spots were.
A timetable's usefulness comes from the fact that the sheet is separate from the train itself. The same timetable can run one train today and a different train tomorrow. A motion table works the same way. Write it down once, and after adjusting for body proportions, it can carry over onto a completely different character, tall or short.
In detail
What to write down gets decided first
Building a timetable means deciding first which stations to time. Every point along the route can't be recorded. Movement works the same way. Instead of recording the whole body, a few dozen bendable spots — the joints — get picked, and more if fingers or the face are included.
The chosen points come with fixed names and order, and which point connects to which gets written down too. That connection information is what lets whoever reads the table later reconstruct how an arm was bent.
What gets picked decides what the table can do. A table that skips finger joints can never reconstruct a fist closing. And once the table is finished, there's no way to know how a missing point actually moved; it can only be drawn in by hand or filled in by guesswork.
One instant is one line in the table
The table doesn't get built all at once — it accumulates one line per instant. From webcam footage, somewhere between thirty and sixty lines a second is typical; studio equipment often writes more than a hundred. Denser lines keep a fast motion looking smooth; sparser lines make a quickly swung arm look choppy and broken.
What happened in the gap between lines isn't in the table at all. Reconstructing the motion means smoothly filling that gap between the line before and the line after. That's why an extremely fast motion sometimes ends up smoothed out more than it really was.
It can be recorded without a special suit
For precise capture, reflective markers get stuck all over the body and filmed inside a room ringed with multiple cameras. Watching a marker's position from several angles at once pins down exactly where that point sits in space. Accurate, but slow to set up and tied to one location.
These days another method has spread widely: it just takes a single ordinary camera feed and guesses point locations from it. Left-right and up-down position on screen come out fairly reliably; how far a point sits front-to-back gets estimated from cues like the body's overall size and proportions.
That's why a table built from a webcam wobbles during a sideways turn or whenever an arm gets hidden behind the body. In exchange, it needs almost no setup, which makes it far more convenient for casual use.
The table sits apart from the body
A finished table carries none of the real person's appearance. That's exactly what lets it get dropped onto a completely different body. Copying raw locations directly causes problems here, though — transfer a wrist position straight onto a character with a different arm length and the arm stretches to reach it.
So what gets transferred is how far each joint bent, rather than the location itself. Hand over how many degrees the shoulder lifted and the elbow bent, and the receiving character builds that same angle using its own arm length. A shorter arm ends up reaching a different spot, but the shape of the motion holds.
Gaps and jumps show up in the table
A recording is never perfectly clean. An arm swinging behind the body leaves a gap where the wrist location is unknown for that instant; a lighting flicker can send a single line jumping to the wrong spot. Play it back as is and a hand suddenly judders.
So a cleanup step always runs before the table gets used. Gaps get filled in from the lines before and after; a single stray value gets smoothed to match its neighbors. Impossible readings, like a bone suddenly changing length for one line, get caught and corrected too.
More precisely
Motion capture records the position of predefined joint points over time. The result is usually stored as a skeleton structure with a rotation value for each joint attached to every frame. Optical setups using reflective markers and multiple cameras, inertial setups using small body-worn sensors that measure tilt and acceleration, and video-based setups that estimate location from camera footage all see use side by side. Moving a finished table onto a different skeleton is called retargeting.
The analogy breaks down somewhere too. A train timetable records position along a single line — a route — but the points in a motion table sit in a space with front-back, left-right, and up-down all at once, along with the angle each one is bent at. Those points are also bound together by bones, a constraint that says their length can't change — a condition a timetable never has to deal with. Above all, a timetable is a plan someone laid out in advance, while a motion table is a record of something that actually happened, measured and kept.
Try it yourself
Common misconceptions
It's easy to think motion capture copies video directly, but actually only the numeric location of predefined points survives — none of the appearance or background carries over.
It's easy to think a special suit and studio are required, but actually a table can be built from webcam footage alone, though it wobbles in sections where the body gets hidden.
It's easy to think a table alone makes any character move instantly, but actually body proportions differ, so a separate step converts to angles and cleans up the result first.
One-line summary
Motion capture turns movement into a location record for a set of predefined points, and because that record sits apart from the body, it can be converted to angles and carried over onto a different character.
Spotted an error or have a better analogy? Suggest an edit · Last updated 2026-10-17