Gracia

Gracia is a video editing tool. Gracia reconstructs real scenes and performances as 4D Gaussian splats for sports broadcasts, film and VFX, and Physical AI. It offers project-based services plus a self-service 4DGS processing suite with SDKs.

Gracia is a 4D reconstruction service that turns real-world scenes and performances into volumetric, time-based 3D data. Its pipelines are built around 4D Gaussian splatting (4DGS), a technique that captures a moving subject so it can be viewed from any angle after the fact. The company positions the output for settings where fabricated detail is unacceptable, since the geometry comes from real capture rather than generative guesswork. The offering is split into four areas. Physical AI and robotics teams use the data for system identification, real2sim2real workflows, evaluation and benchmarks. Entertainment customers use free-viewpoint replays for sports and live events, and the site points to a 2026 PGA Championship broadcast that used 4DGS for golf swings. Film and VFX teams get real performers with accurate composition and freedom to change camera and framing in post, illustrated by a case study on Muse's Nightshift Superstar video, where the singer was turned into a 4D Gaussian splat. The fourth area, 4DGS Processing, is described as a self-service reconstruction suite with SDKs for distribution. In practice, most of the work is project based. Visitors choose a project type (sports and live events, VFX and filmmaking, Physical AI and robotics, 4DGS processing or other) and send an inquiry through a form, after which the team scopes the engagement. The self-service suite is the exception, aimed at teams that want to run reconstruction themselves and ship results through SDKs. The homepage does not publish pricing, capture hardware requirements or technical specifications, so those details have to be requested. Among alternatives, Gracia sits closer to volumetric capture studios and reconstruction toolchains than to text-to-video generators. Tools such as generative video models invent footage, while Gracia reconstructs what was actually recorded. That makes it relevant to broadcasters, VFX studios and robotics researchers who need fidelity and viewpoint control, and less relevant to casual creators looking for quick clips.