3D BLOG/ARTICLES/FIRST PHOTOGRAPHY, THEN VIDEO, AND NOW… GAUSSIAN SPLATTING?

First Photography, Then Video, And Now… Gaussian Splatting?

May 18, 2026

Article by Michael Rubloff

Gaussian splatting is a groundbreaking imaging technology - first published in 2023 by French research group INRIA - that extends visual capture from 2D into strikingly lifelike 3D. It is already transforming industries from filmmaking to robotics, and it may ultimately succeed photography and video as the dominant form of visual expression.

As the founder and managing editor of RadianceFields.com, I have been writing almost daily about lifelike 3D for a little over three years, staying on top of research, hardware, software, and real-world deployment. Having witnessed the evolution and adoption of Gaussian splatting firsthand, I believe we are in the presence of a foundational imaging medium, one that deserves the same cultural attention we once gave to the invention of the photographic camera.

What Is Gaussian Splatting?

Gaussian splatting is a form of radiance field representation - an alternative to the more traditional photogrammetry approach - that gained traction immediately after its introduction. One key reason: it offered significant advantages over earlier radiance field technologies like NeRF (Neural Radiance Fields), introduced in 2020, including dramatically faster rendering speeds.

A defining feature is its support for view-dependent effects. In plain terms: when you look at a reflective surface like glass or a mirror in the real world, the light shifts as you move around it. Gaussian splatting replicates that behavior with remarkable accuracy, something earlier technologies consistently struggled with.

Those who had existing datasets or hand-modeled 3D assets could reconstruct them with far greater realism. And even a series of standard multi-view 2D images or video could serve as input, reconstructed as static or dynamic 3D. In short: Gaussian splatting made lifelike 3D accessible to anyone who was already capturing the world in two dimensions.

Why It Matters: A New Foundational Medium

Photography has existed for roughly 200 years, and moving images for well over a century. There is no reason they should be the final form of how we capture reality.

Gaussian splatting keeps growing, and it will continue getting better and more popular. I would go so far as to propose that radiance fields are a technology with the potential to succeed traditional photography and video as the dominant form of visual capture and expression. That may sound like a bold claim, but the evidence across industries and applications is certainly mounting.

Where It's Being Used

Film and Entertainment

One of the most visible recent examples of Gaussian splatting in action is James Gunn's 2025 blockbuster Superman. A key plot point involves a 3D hologram-like message recorded by the hero's Kryptonian parents - played by Bradley Cooper and Angela Sarafyan - and later restored to full clarity.

To create it, both actors were brought to Infinite Realities, a company in Ipswich, UK, where roughly 200 cameras captured their performances from every possible angle. That data was then reconstructed using Gaussian splatting into complete dynamic 3D, allowing the post-production team to determine camera movement, angles, and focal lengths long after the performance had wrapped.

If you look closely at the scene where the degraded hologram breaks apart, you can see the rough edges of the technology itself, left in intentionally because it felt like a natural way for a dynamic reconstruction to dissolve. The medium, as they say, became the message.

Beyond narrative filmmaking, Gaussian splatting is transforming how productions operate day to day. For location scouting, instead of flying multiple employees to destinations around the world, companies can send a single person with an iPhone, a 360 camera, or a SLAM-based camera built specifically to capture Gaussian splats efficiently. For pre-visualization, studios can hand reconstructed environments to directors and production teams, who use software like Unreal Engine to plan shoots - camera placement, blocking, logistics - without ever physically visiting the location.

Architecture, Engineering, and Construction

Some of the earliest adopters of Gaussian splatting were in architecture, engineering, construction, and the geospatial industry. These sectors had long relied on photogrammetry, but that approach struggles with thin structures and reflective materials, including everyday elements like foliage and water.

Gaussian splatting handled these out of the box. Simply put, these industries could now clearly visualize environments they had already captured but could never quite see before.

Robotics and Simulation

One of the strengths of Gaussian splatting is producing remarkably lifelike reproductions. So much so that robotic systems frequently cannot distinguish them from the real world, which makes them extraordinarily valuable for simulation and training.

Companies can capture a city, reconstruct it digitally, insert synthetic vehicles, and simulate traffic patterns without deploying a single vehicle in the real world. NVIDIA's Alpamayo platform, announced at CES, uses a variant called 3D Gaussian Unscented Transform as part of its AlpaSim product. Third Dimension's SuperSim ingests LiDAR and RGB image data to rapidly generate simulation environments. In other words: robots can now be trained in virtual worlds that are nearly indistinguishable from the real one.

Virtual Reality

The VR space is already shifting toward Gaussian splatting. Meta's Hyperscape Capture on Quest 3 and 3S allows users to capture scenes directly with the headset; Meta then processes and renders them in real time with striking fidelity, far more detailed than the cameras involved would suggest. These environments can be shared with up to eight people simultaneously, opening meaningful possibilities for social connection, construction planning, and remote collaboration.

What's Next: The Future of Gaussian Splatting

Looking forward, I believe any industry currently using 2D imaging will eventually shift toward 3D. Insurance, e-commerce, archaeology, education: at a minimum, anything that touches 2D imagery stands to benefit. It is simply a better way to interact with the physical world, and even with synthetic ones.

I also think Gaussian splatting paired with AI and large language models is increasingly likely. Imagine asking a system to walk you through the assembly of a disassembled object, or moving through a museum reconstruction that responds to your questions in real time. Virtual tours and interactive educational content built on radiance fields are not far off.

At this point in technological development, photogrammetry and radiance fields sit alongside one another rather than replacing each other. We do not yet know whether NeRF, Gaussian splatting, or something we haven't seen yet will define the next era of imaging.

Personally, I am agnostic. I am mainly excited that lifelike 3D now exists and that its evolution continues picking up speed.

‹ BACK TO ARTICLE LIST

on offSubscribe