A Leap Forward at Movement Day
Our second-generation spatial motion models harness AI and physics to push 3D motion capture beyond limits—from a single camera to massive scales. This isn’t just an upgrade; it’s a revolution for creators, researchers, and industries.
5 new models
s2
(highest quality)
s2-light
(faster processing)
m2
(highest quality)
m2-xl
(extra large spaces)
enhance
s2 and s2-light models
s2 is the most accurate single-camera model we’ve ever built. Compared to its predecessor s1, s2 features major improvements in depth estimation from a single view, leading to: Greater motion stability, significantly better foot planting, smoother and more coherent motion reconstruction, track multiple actors from a single camera. Single-camera motion capture is inherently challenging due to the lack of depth information. s2 sets a new benchmark by intelligently inferring 3D motion with unprecedented fidelity — all from a single camera feed.
s2-light is a streamlined version of the s2 model, specifically optimized for maximum inference speed (up to 20x faster than s2) while maintaining good-quality motion reconstruction with only minimal trade-offs. Ideal for: Fast prototyping, storyboarding, near real-time applications where speed is critical. If you need a quick, efficient understanding of human motion, s2-light is the perfect solution.
Greater motion stability
Improved foot planting
Smoother and more coherent motion reconstruction
Track multiple people
m2 model
m2 is our most powerful multi-camera model yet, surpassing m1 in every way. With m2, you can achieve optical-grade motion capture quality at a fraction of the cost — even in challenging environments.
Enhanced spine biomechanics
New 6 degrees of freedom shoulder model
Smoother and more stable motion for complex and dynamic scenes
m2-xl model
m2-xl takes multi-camera motion capture to the next level, scaling m2’s capabilities to massive environments. It supports large volumes and multi-person tracking — capturing over 20 people simultaneously. Operates in spaces as vast as a football stadia (with appropriate camera setups). Optimized and stardarized techniques for accurately calibrating RGB cameras in large volumes. Whether you're working on large-scale productions, live sports, or immersive experiences, m2-xl unlocks unprecedented freedom and flexibility in motion capture.
Superior handling of high-occlusion environments with multiple interacting subjects
Track over 20 people
Works in extra large spaces
enhance model
We’re thrilled to introduce our first motion-to-motion model, designed to give you more control and flexibility over motion capture data than ever before. Whether you’re refining noisy motion or dealing with challenging occlusions, this model unlocks powerful new capabilities from automated cleanup to plausible motion generation for highly occluded scenarios.
