Cartwheel

Cartwheel documents Mark Andrews' Dogstar team completing a one-week production sprint using phone reference video, Comic3 motion retargeting on existing character assets, and Unreal Engine rendering for episodic social content.

Cartwheel reports Psyop using Comic3 to reconstruct unstable body motion from original mocap footage for a two-character commercial, retaining original head/hand details and delivering final spots without reshooting.

Cartwheel applies generative AI to 3D animation workflows, but public evidence of large-scale enterprise deployment remains limited.

Cartwheel's August 2026 research release describes training hundreds of human-motion models at varying scales and studying compute-optimal scaling for both autoregressive and flow-matching motion generation, with a linked public research repository.

Cartwheel develops generative models and tooling for controllable 3D character animation.

Cartwheel develops generative models and tooling for controllable 3D character animation.

Cartwheel names its organizational roles across principal research, generative vision, computer vision, data engineering, animation engineering, product and operations, alongside CEO Jonathan Jarvis and chief scientist Andrew Carr.

Craft Ventures states it led Cartwheel's $10 million seed round in May 2025 with WndrCo, Tirta Ventures and Runway, and participation from existing investors including Accel, Khosla and Human Ventures.

Cartwheel has built a specialized AI animation product and research team, but remains an earlier-stage organization.

Cartwheel documents motion representations including 77 joints, facial blendshapes and motion derivatives, with FBX, glTF, USD, BVH and MB outputs and multiple frame rates.

Cartographer documents rig recognition, body and facial retargeting, appendage spring physics, collision offsets and MuJoCo robot-data export through an API.

Generative modeling and graphics skills transfer to adjacent creative AI roles, while 3D animation specialization limits breadth.