Motion capture records a performer's movement and applies it to a digital skeleton. It is fast for realistic human motion and it is not free of animation work.
What it is good at
- Realistic weight and timing. The hardest things to hand-animate convincingly.
- Volume. Hours of usable locomotion, idles and interactions in a session.
- Performance. An actor's intention carries through in a way keyframing struggles to reach.
What it is bad at
- Stylisation. Exaggerated, cartoon or superhuman motion is a hand-animation job.
- Anything a human cannot do. Creatures, impossible physics, non-humanoid proportions.
- Precision. Exact contact with virtual objects that were not physically present during capture.
The cleanup
Raw capture data is never final. It arrives with:
- Jitter, from marker noise or solver error
- Foot sliding, corrected with IK
- Interpenetration where the performer's proportions differ from the character's
- Missing or occluded markers, producing gaps that need filling
Budget for cleanup as a real stage. The rule of thumb across studios is that capture shortens animation time substantially without eliminating animator involvement.
Retargeting is where the constraints bite
Capture data is skeleton-specific and applied to characters through retargeting. A character with a standard humanoid skeleton can use a motion library directly; one with an idiosyncratic skeleton needs mapping work per clip.
This is the strongest practical argument for auto-rigging producing a conventional skeleton: it is what makes existing motion libraries usable without a per-character setup.
Alternatives
Video-based capture without markers has become genuinely usable for many purposes, and lowers the cost of entry considerably. It trades some precision for not needing a stage, a suit or a marker system.