By Patrice Rondao Alface, Principal Standardisation Specialist, Nokia
Just as imaging technology unlocked the world of digital photography for 2D images, Gaussian splatting is poised to do the same for 3D – unlocking photorealistic volumetric scenes that can be rendered and interacted with in real time. Within two years of its introduction in a 2023 research paper, Gaussian splatting has moved from academic experimentation to a viable production tool redefining how we capture, reconstruct, and experience the physical world in digital form.
Consider a film or television production where location environments are recreated digitally. Every surface, shadow, and reflection rendered with the realism of live footage, yet adjustable in real time for lighting or camera movement. Gaussian splatting makes this kind of scene capture and manipulation possible, offering a bridge between the flexibility of virtual production and the authenticity of natural cinematography.
Traditional 3D content production pipelines have long faced a trade-off between realism, speed, and flexibility. Gaussian splatting changes that balance entirely. By allowing every viewpoint in a scene to be explored interactively – with accurate reflections, lighting, and texture behaviour – it brings real-world physical and visual complexity into the digital domain with unprecedented efficiency. It represents a shift comparable to the one that transformed 2D imaging a generation ago.
The new frontier of 3D reconstruction
At its core, Gaussian splatting rethinks how digital scenes are represented. Instead of relying on traditional polygon meshes or computationally heavy neural radiance fields (NeRFs), it uses “splats” – 3D points that carry information about colour, light reflectivity, opacity, volume and orientation. These splats can be quickly adjusted and displayed, enabling interactive, photorealistic 3D scenes that respond instantly to changes in viewpoint or illumination.
What sets this approach apart is its balance between fidelity and efficiency. Advances in compression and processing mean these photorealistic 3D models can run smoothly on everyday devices, not just high-end workstations. This accessibility has accelerated experimentation across the content ecosystem – from game developers and film studios to research labs – all using splatting to render richer scenes faster, cut down post-production steps, and bring real-time realism closer to everyday devices.
Industries set to benefit
Gaussian splatting’s combination of realism and responsiveness makes it relevant across any sector where spatial accuracy and immersion matter.
In gaming and film, Gaussian splatting opens the door to cinematic-quality worlds that render in real time – vast landscapes and lifelike characters that respond instantly to light, motion, and perspective. In retail and real estate, customers will be able to step into virtual showrooms and explore products or properties from any angle, as if physically there – something that has been typically difficult to do, due to frequently inconsistent lighting and illumination effects in real estate imagery. In engineering, cultural heritage, and education, richly detailed digital twins will let teams collaborate inside virtual replicas of factories, aircraft, or archaeological sites, examining and modifying them in real time from anywhere in the world.
Gaussian splatting can also run efficiently on everyday devices – from laptops to smartphones. It takes just a few minutes to apply splatting to pictures from different angles, making 3D visualisation more accessible. This is paving the way for a future where creating and experiencing 3D content truly becomes as natural as taking a photo.
The road ahead
Static and even dynamic Gaussian splatting, where movement and animation are rendered in real time, is already possible. The next step is scaling that progress – embedding these capabilities into creative workflow and consumer platforms so they become part of everyday content creation. Achieving this will rely on close collaboration between researchers and creators to ensure efficiency, interoperability, and responsible development as photorealistic 3D moves from breakthrough to mainstream reality.
Just as digital photography redefined how we record 2D memories, Gaussian splatting could become the universal medium for 3D – realistic, dynamic and widely available. The transition is already underway. The opportunity now sits in turning its rapid progress into practical, scalable applications that transform how we create and connect through immersive digital media.
Image caption: ‘Model represented on the left with a point cloud and on the right with Gaussian splats’.
