Can 3D Gaussian Splatting replace a laser scan for architects?

For most everyday uses (estimating surfaces, distances, spatial organisation), yes. For millimetre-accurate surveys intended for an architect or engineering firm, a point-cloud laser scanner remains the reference. The two technologies can be combined on the same project: laser scanning for technical drawings and 3DGS for communication, archiving and experience.

What is the difference between 3DGS and photogrammetry?

Both reconstruct a 3D space from photographs, but the result is very different. Photogrammetry produces a polygonal mesh with applied textures, heavy computation, and limited visual fidelity on complex materials (glass, water, vegetation). 3DGS produces a cloud of Gaussians that renders light and materials photorealistically, with real-time rendering in a browser. 3DGS is faster to produce and visually superior for communication.

Who can view a 3DGS model, do you need a VR headset?

No headset or app is needed. A 3DGS model is viewed in a standard web browser, on a computer, tablet or recent smartphone. You just click a link. A GPU capable of real-time 3D rendering is required, which is the case for almost every device released in recent years.

Where does 3D Gaussian Splatting stand in 2026?

In two years, 3DGS has moved from research paper to production. Workflows are stabilised, web viewers mature, and the first commercial offerings structured. The technology remains emerging on the market, which creates a window of opportunity for companies adopting it now. The next developments concern compression, streaming, and gradual integration with BIM and AR/VR workflows.

3D Gaussian Splatting: the complete guide to the technology redefining 3D

3D Gaussian Splatting, or 3DGS for short, has moved in less than three years from the research lab to professional production. If you work in communication, heritage, construction, events or healthcare, you will encounter this technology more and more often. This guide aims to give you a clear understanding of what 3DGS is, what it does better than the alternatives, its real limits, and how to use it in a concrete project.

The approach is deliberately educational. No unnecessary jargon. When a technical term appears, it is explained.

What is 3D Gaussian Splatting?

3D Gaussian Splatting is a 3D capture and rendering technology that reconstructs a real space from photographs, and renders it photorealistically, freely navigable, in a simple web browser.

Concretely, after a capture lasting a few minutes to a few hours, you get a 3D model that you can explore as if you were there, from every angle, with no headset, no app, no training. You share a link, the person clicks, they navigate.

A simple analogy

Imagine the difference between a paper map and Google Street View. The map gives the information, but not the experience of being there. 3DGS is to classic 3D what Street View is to the map. It is the same spatial information, rendered experientially rather than representationally.

The fundamental difference with a classic 3D mesh

Classic 3D modelling (a polygonal mesh) reconstructs an object or a space by assembling triangles, onto which photographic textures are applied. It is effective for objects with simple shapes and uniform materials, but struggles with complex materials (glass, mirrors, vegetation, fabrics, human skin), with light, and requires a great deal of time in manual post-production.

3DGS abandons that logic. Instead of reconstructing surfaces, it reconstructs the light that reaches the capture points, placing millions of small Gaussian distributions in space (hence the name), each carrying a colour, a transparency and an orientation. The final render is not an assembly of polygons; it is an accumulation of these Gaussians seen from your current viewpoint.

The result: photorealistic fidelity that includes light, reflections and difficult materials, computed in real time on a standard GPU.

How does 3DGS work?

You do not need to understand the mathematics to use the technology, but grasping the broad strokes helps you specify it better on a project.

Step 1, the capture

On site, a camera (or a specialised device such as the PortalCam) is moved in continuous motion to cover the space. Several hundred to several thousand images are taken from many angles. No fixed points, no tripod every two metres. Motion is the key.

Step 2, the alignment

Software determines the 3D position of each image in space by analysing the overlaps between photos. This is called Structure from Motion. At this stage, you have a 3D point cloud that serves as a base.

Step 3, the optimisation of the Gaussians

The 3DGS algorithm places 3D Gaussians in this space and iteratively adjusts their position, colour, orientation and transparency so that, seen from each capture image, they reproduce as closely as possible what was photographed. This step is computed on the GPU, in a few minutes to a few hours depending on the complexity of the scene.

Step 4, the real-time rendering

The final model is exported in a lightweight format (.PLY, .SOG, or the viewers’ proprietary variants), loaded into a web viewer, and rendered in real time in the browser. It is the browser that computes the images as you move the mouse or finger.

See the process in video

To make all this concrete, here is a short video showing the alignment of the images, the optimisation of the Gaussians during training, and the final navigable result.

It is the combination of these four steps that makes the technology economically viable for professional projects. None is new in isolation. It is their combination, formalised by an INRIA paper in 2023, that changed everything.

Why 3DGS is a game changer now

Photorealistic 3D capture is not a new subject. What is new is that it has become accessible.

Before 2023

To produce a photorealistic 3D model of a place, two options existed. Either manual modelling by a 3D studio, weeks of work, a high budget, a fixed result. Or classic photogrammetry, heavy computation, an imperfect mesh on complex materials, a final file weighing gigabytes, hardly navigable in real time.

360° virtual tours filled the gap by democratising a simplified format, but with navigation limited to fixed points.

The technological leap

3DGS, formalised in late 2023, brought three simultaneous breakthroughs. First, photorealistic rendering that correctly includes light and difficult materials. Second, a computation speed that allows a model to be produced in hours rather than days. Third, real-time rendering on a consumer GPU, so it can be viewed in a browser on a phone.

2025 and 2026, the tipping point

In 2025, the first commercial offerings emerged, mainly in the United States and Asia. In Europe, and particularly in Switzerland, adoption begins in 2026. This is precisely the moment when the technology becomes usable predictably on client projects, with stable workflows and mature viewers.

For companies, it is an early-adoption window. Not too early (the technology is mature), not too late (your competitors have not all seized it yet).

Concrete use cases

3DGS applies wherever a real space needs to be documented, shared, experienced or archived. Here are the fields already using it in production.

Heritage, culture, tourism

Digitisation of historic monuments before restoration, a 3D reference archive for heritage authorities, augmented cultural mediation for museums (guided paths, enriched narration), accessibility for distant or reduced-mobility audiences, online tourism promotion.

Construction, real estate

Site monitoring through recurring captures to compare progress at different dates, probative 3D documentation in the event of a dispute, sharing with remote project owners and subcontractors without travel, archiving before renovation, high-end real estate promotion. Our 3D site monitoring guide details the workflows.

Events and MICE

Immersive presentation of a seminar or gala venue before booking, sharing a specific configuration (seating, stage, signage) with sponsors, post-event continuation to reuse a stand or a set, sales enablement for teams selling a venue remotely.

Healthcare and education

Pre-operative patient pathways to reassure (paediatrics, maternity), immersive presentation of a premium room for remote conversion, modern institutional communication, virtual open days for universities, presentation of the laboratories and workshops of a technical school.

Industry

Documentation of technical installations, operator training and onboarding, asset management, quality and safety control, knowledge transfer before an expert leaves, a digital twin of a production site for predictive maintenance.

What 3DGS does better than the alternatives

Four alternatives exist today to digitise a space. Understanding their respective strengths and weaknesses clarifies when 3DGS is the right choice, and when it is not.

Versus 360° virtual tour

360° navigates from fixed point to fixed point, by teleportation. 3DGS navigates freely through the whole space. On large volumes or complex spaces, the difference is immediately visible. Our detailed 3DGS vs 360° comparison goes into the detail.

Versus point-cloud laser scanning

Laser scanning produces millimetre metric surveys, the reference base for architects and engineering firms. 3DGS produces a photorealistic experience, the reference base for communication and archiving. The two are complementary more than competing. You can combine a laser scan for technical drawings and a 3DGS for communication on the same project.

Versus classic photogrammetry (mesh)

Photogrammetry reconstructs polygons with applied textures. It struggles with complex materials (glass, vegetation, skin, water, changing light). 3DGS handles these materials naturally. On production speed, 3DGS is also faster. The mesh keeps an advantage when you need an object exportable for a classic 3D pipeline (video game, animation, 3D printing).

Versus manual 3D modelling

A 3D modeller can create an idealised representation of a place, perfect and stylised. That is useful for conceptual architecture, projects not yet built, or marketing visualisations. 3DGS captures the real, as it is, with its imperfections. These are two different uses; they do not compete.

Our article building digital twin, 4 technologies compared offers an advanced technical comparison.

The real limits of 3DGS

No technology is perfect. Here is what 3DGS does not do, or does less well than the alternatives.

Metrological precision

Measurements extracted from a 3DGS model are usable for most everyday purposes (estimating surfaces, distances, spatial organisation), but are not guaranteed to the millimetre. For surveys intended for an architect or engineering firm, a laser scan remains the reference.

Performance on older devices

Real-time rendering requires a GPU capable of handling 3D in a browser. That is the case for all recent computers, high-end tablets and smartphones less than 4 to 5 years old. On very old or very low-end devices, rendering may be choppy. For an audience that must be reached including on old hardware, 360° remains more universal.

Maturity of the software ecosystem

3DGS web viewers exist (XGRIDS LCC Viewer, PlayCanvas, SuperSplat, and others), but the ecosystem is less mature than that of classic 3D formats. Exports to other tools, and integrations with BIM pipelines or game engines exist, but still require a degree of expertise. It is a limit that fades month after month, but it exists in 2026.

Cases where another technology remains relevant

If your goal is to produce technical drawings for a building permit, use a laser scan. If your target mostly uses old devices, use 360°. If you need a model exportable to a standard game engine, photogrammetry remains simpler. 3DGS is not the answer to every question. It is the best answer to many of them.

How long does it take to produce a 3DGS model?

Three steps, three durations.

On-site capture, from a few minutes for a small room to a few hours for a large volume. A cathedral, one to two hours. An apartment, twenty minutes. An industrial site, half a day to a day depending on the extent.

Processing, a few hours of GPU computation, in parallel with other tasks. For a client project, processing and QA are planned over 24 hours.

Delivery, the final model is delivered within 24 to 48 hours from the end of the capture. It is a standard that few other 3D visualisation technologies can reach.

For comparison, a classic 3D model of equivalent quality generally takes two to six weeks.

How to get started with 3DGS in Switzerland?

Four criteria matter when choosing a provider.

1. Mastery of the full pipeline

3DGS is not just a capture. It is a workflow that includes scouting, capture, alignment, optimisation, delivery on a suitable viewer, and hosting. Check that the provider masters the whole chain, not just the shoot.

2. The viewers offered

Several web viewers exist and each has its strengths. A good provider chooses according to the project, not by default. For a museum, the viewer must handle narration. For a construction site, the history of captures. For a marketing site, iframe integration.

3. Long-term hosting

A 3D model must remain accessible in 5 years, not just 5 months. Understand who hosts it, where, at what recurring cost, and what happens if the relationship ends.

4. The ability to propose fallbacks

3DGS remains emerging. On certain sites (difficult light, access constraints, highly reflective surfaces), a fallback deliverable may be necessary. A good provider knows this, and proposes options (photos, 2D drawings, complementary 360°).

See our three engagement packages to understand how we structure projects, from raw capture to a bespoke interactive experience.

The future of 3DGS, where the technology is heading

Three directions are taking shape for 2026 and 2027.

Compression and streaming

Today, a good-quality 3DGS model weighs from a few tens to a few hundred megabytes. Compression work aims to divide this weight by 5 to 10, to enable on-the-fly streaming and viewing over modest mobile connections.

Mobile integration

Native iOS and Android viewers are advancing. Rendering on smartphones is becoming smoother, paving the way for AR (augmented reality) uses where a 3DGS model is integrated into the physical world seen through the phone’s camera.

BIM and industrial workflows

The integration between 3DGS and BIM is being structured. 3DGS brings the realistic, navigable layer; BIM brings the structured, metric layer. The two will enrich each other rather than compete.

Over a 2 to 3 year horizon, we can expect 3DGS to become the default format for immersive communication of real spaces, while classic 3D mesh remains for creation (games, animation, simulation), and laser scanning for technical precision.

In summary

3D Gaussian Splatting is the technology that finally makes photorealistic 3D capture viable for professional projects, in short timeframes, at controlled costs, and accessible from a simple web link. It does not replace laser scanning on precision, nor 360° on maximal compatibility, but it outclasses manual modelling and classic photogrammetry on the majority of business cases.

In 2026, adoption is beginning in Europe and in Switzerland. It is the right time to integrate the technology into a communication, documentation or archiving strategy, before it becomes a commonplace standard.

To go further, see our 3D Gaussian Splatting projects, compare the available technologies in our 4-technology comparison, or discover the use cases by sector via our heritage, construction, events and healthcare pages.

Have a project in mind? Let’s talk about it; we will give you an honest recommendation based on your goals.