Volinga Adds 4D Gaussian Splatting to Its Unreal Engine Pipeline

Volinga has added 4D Gaussian Splatting support to its Unreal Engine workflow, letting animated splat captures play back in real time inside the engine.

The update extends Gaussian splatting into the time dimension. Where a standard 3D Gaussian Splat reconstructs a static scene as a cloud of points, 4DGS captures motion, so a splat can hold a moving subject or a changing environment rather than a frozen moment.

• A PLY sequence converts to a single NVOL asset. The Volinga Suite processes an animated capture and outputs one optimized file for the engine.
• The Volinga Plugin imports that NVOL and plays it back in Unreal. Dynamic splats run with native engine controls.
• UE 5.8 support ships alongside DLSS, sublevel streaming, and Sequencer integration.

The Volinga Suite turns a PLY sequence into one NVOL file that plays in real time

The pipeline splits across two tools. The Volinga Suite handles capture processing and conversion, taking a sequence of PLY files that each represent a single frame of animation and encoding them into one NVOL asset. The Volinga Plugin then imports that asset into Unreal for playback.

NVOL is Volinga's own extension of the PLY format. According to a Digital Production interview, the format carries additional metadata that PLY was never designed to hold, which is what lets the engine treat a splat as a native asset. The plugin places splats in Unreal's content browser and lets artists work with them the way they would any other object in the scene.

Collapsing a per-frame PLY sequence into a single file addresses the practical problem with animated splats: a moving capture is a stack of separate point clouds, and playing that back frame by frame is heavy. One encoded NVOL asset gives the engine a single object to stream and render.

UE 5.8 support arrives with DLSS, sublevel streaming, and Sequencer integration

Beyond the 4DGS conversion, the release brings the plugin current with Unreal Engine 5.8 and adds several engine-side features for working with splat scenes:

• DLSS support for upscaling splat renders.
• Sublevel streaming so large splat environments load in sections rather than all at once.
• Sequencer integration to place and time splats on Unreal's timeline for shots.
• Exposed Radiance Fields properties that give artists direct control over the splat inside the engine.
• Improved VRAM efficiency and large-file support for heavier captures.

Sequencer and sublevel streaming matter most for virtual production and previs, where a splat has to behave like a normal set piece across a timeline and inside a large environment rather than sitting as a static prop. We previously covered how Volinga Plugin Pro brought production-ready 3DGS to Unreal, including relighting and color controls; the new engine features build on that plugin.

SPZ and SOG compatibility plus user masks widen what the Suite can ingest and clean

On the capture side, the Suite adds compatibility with the SPZ and SOG splat formats, so scenes trained or exported outside Volinga can move into the pipeline. User masks give artists a way to isolate or exclude regions of a capture during processing, useful for removing unwanted elements before an asset reaches the engine.

The format support fits a workflow Volinga has been building toward. Its partnership with XGRIDS routed LiDAR-based real-world scans into the Unreal pipeline, and broader format compatibility means fewer captures get stranded in an incompatible export.

Motion capture covers the case static splats could not

Static Gaussian splats already gave VP teams a fast route from real-world capture to a real-time Unreal environment. Adding motion covers the case those static scenes could not: a performer, a vehicle, water, or any subject that has to move inside the shot. Whether 4DGS holds up at feature quality depends on capture conditions and scene complexity, and Volinga has not published pricing or hardware requirements for the dynamic workflow. For teams already running the plugin, the update lands as an extension of a pipeline they know rather than a new tool to learn.