Imagine this: You're watching a long video, maybe a concert or a sporting event. Now, imagine that video isn't just a flat image, but a 3D recording where you can look around and see the scene from any angle.

That's a "volumetric video," and storing them can be a huge challenge.

This research shows how AI can make these 3D videos much smaller and easier to handle.

Think of it like this: the video isn't just one big chunk of data. Instead, the AI breaks it down into smaller pieces and figures out which parts are changing quickly and which parts are staying the same. This allows for much more efficient storage.

The result? You can enjoy longer 3D videos with the same amazing quality, without needing massive amounts of storage space. This could lead to incredible new experiences in virtual reality and beyond.

This research ''Temporal Gaussian Hierarchy'' was conducted by a team from Zhejiang University, Stanford University, and HKUST, and is published in ACM Transactions on Graphics.

Credit: Endrit Restelica