First 4D Gaussian Splat integration in live broadcast

In sports, the action happens in milliseconds. A championship swing is over before a director can cut to it, and whichever camera was live is the only story anyone gets. Everything above rides on cameras that can each show one fixed angle.

And a live championship happens once. Whatever the cameras miss is lost forever, and a story that arrives after the broadcast has moved on never reaches the fans at all.

The story. The performance is over in the blink of an eye. Telling it with coherence and clarity is the whole product of a live broadcast.

The understanding. Athletes, coaches, and broadcasters need depth plus time, unlimited views, to study and analyze what actually happened.

The moment. One take, no second chances. Whatever is not captured cannot be replayed, taught, or shared.

Today's answer is more cameras. But more cameras only means more cuts: every angle is still a flat, fixed viewpoint, and stacking them adds footage, not coherence.

And the traditional 4D pipeline that could solve it, download, trim, align, color, transcode, upload, consumes hours of operator time for 2 seconds of content. It works for a studio. It cannot keep pace with live.

More cameras, more cuts, no coherence. Each added camera is another fixed viewpoint to cut between. None of them lets the viewer, the coach, or the editor actually move around the swing.

The 4D pipeline was built for studios, not live. Offloading, trimming, aligning, color-matching, transcoding, and uploading 56 cameras takes hours per key moment, multiple machines, and on-site servers.

A championship gives you minutes, not hours. Four athletes to capture among 20 broadcasters and 200,000 fans, the entire stage and crew on and off the field in minutes, 5 minutes to capture the environment, and the finished camera moves due back within hours. Trucked-in infrastructure would have made it impossible.

Meridian 4D wrapped each golfer in 56 synchronized iPhone 17 Pro cameras, a wireless, battery-powered stage that repositioned from tee box to fairway to green as the story moved across the course. Nothing is generated or approximated: every frame comes from real synchronized capture of the real athlete in the real environment. 100 percent real.

The camera stops being a fixed position. It becomes a capture volume production can move, reframe, and render from any angle after the swing has already happened.

The breakthrough is on the device. The Meridian App turns every phone into an automated edge computer: each one trims, color-grades, transcodes, and uploads its own footage on-device, in parallel, straight to the cloud for 4DGS processing. No card offloads, no workstations, no trucked-in infrastructure: the entire multi-step pipeline collapses into a single automated step per camera, and hours become minutes.

T-Mobile's untethered 5G made it live. The network connected all 56 iPhones at the same time for remote control, monitoring, and parallel upload, with no fiber or camera cabling anywhere on the course. Gracia AI rebuilt each capture into a navigable 4D Gaussian Splat, ready for the broadcast in hours and explorable from any angle at any speed.

This is what human performance analysis has always been missing. A replay shows one flat view of a swing. A 4D Gaussian Splat is an exact digital double of the athlete and the environment, so athletes, coaches, and biomechanists can orbit the swing, freeze it at impact, slow it to any speed, and stand anywhere, including where no camera could physically be. The performance itself becomes the study material, not one camera's version of it.

The same principle applies to any phenomenon that happens too fast, or only once, to understand as it happens: a dancer's leap, a surgical technique, a crash test, a safety incident. Capture the event in 4D, and it can be examined from every perspective, at every speed, for as long as it takes to learn from it.