Heritage
Immersive walkthroughs
Visual documentation of buildings, interiors, remains, excavation sites and singular spaces, for public interpretation or remote review.
Service · Advanced visualisation
3D Gaussian Splatting (3DGS) reconstructs a real space from photos or video and represents it as millions of gaussians with colour and transparency, so it looks almost like a photograph and can be explored in real time. We carry out the capture on site, process the splat and deliver it published in VisionHub, our web viewer, so heritage bodies, architects, industry and construction can share it with a link. Almost nobody in Spain offers this commercially yet — and for a client reviewing an asset remotely, it is the difference between reading a report and seeing the place. We capture in Spain and across Europe.

Gaussian Splatting (3DGS) is a recent technique that reconstructs a space from photographs or video and displays it photorealistically and navigably in the browser, close to being there. A point cloud is unbeatable for measuring; Gaussian Splatting is unbeatable for seeing and explaining. Together they give you a digital twin that is both accurate and easy to show to somebody who is not an engineer.
Heritage
Visual documentation of buildings, interiors, remains, excavation sites and singular spaces, for public interpretation or remote review.
Industry
Natural visualisation of plants, plant rooms, tanks and service routes, where a navigable photograph removes ambiguity.
Construction
Monitoring, presentation and visual sign-off of spaces before, during or after an intervention.
The right approach is to choose the data by the decision: LIDAR and BIM for metric accuracy, Gaussian Splatting for visual experience and spatial understanding.
Traditional photogrammetry produces point clouds from imagery, from which digital terrain models (DTM), digital surface models (DSM), orthophotos, meshes and metric point clouds are derived — accurate, but without the hyper-realistic quality good visualisation needs. That is where Gaussian Splatting wins: you can move through the scene in high definition with realistic lighting, though you cannot take measurements from it. The two are combined to hold metric data and visual experience at the same time.
They are not substitutes: photogrammetry or LIDAR supply the metric data, Gaussian Splatting the hyper-realistic visual experience. Together they cover measurement and communication at once.
Anyone researching 3DGS soon runs into these names. They are different tools for different stages of the work, and it helps to know what each one does before deciding whether to do it in-house or commission it.
Delivery
This is where we deliver every project. VisionHub opens the splat in PLY or SPZ format straight in the browser and makes it ready to share with a link, alongside the point cloud and the BIM model if the project has them. Whoever receives it installs nothing and does not need to know what a splat is.
Editing
SuperSplat is a free, open-source web editor developed by PlayCanvas. It is used to inspect, clean up, crop and optimise a Gaussian splat that has already been generated. It does not train the model from photos: it works on the finished splat.
Training
Postshot, by Jawset, is desktop software that trains a Gaussian Splatting model from photos or video. It runs on Windows with an NVIDIA GPU, has a free version for non-commercial use and requires a paid licence for professional use and for exporting the splat.
The software is the easy part. What decides whether a Gaussian splat looks good is the capture: coverage, light, overlap and route. That is why our service starts on site and ends with the splat processed, reviewed and published in VisionHub. If you also need to compare viewers for BIM models, our guide to BIM viewers covers it.
A good 3DGS result depends on planning: route, lighting, coverage, overlap, critical areas and the final visualisation objective.
A point cloud stores millions of points with their exact position in space: every point holds real coordinates, so distances, heights and areas can be measured — but up close you see gaps, with no texture and no highlights. Gaussian Splatting turns each point into a soft blob of colour with shape and transparency: volume, reflections and materials appear, and the scene looks almost like a photo, though you can no longer measure on it. In one project both are used: the point cloud to work with, the Gaussian splat to show.
Gaussian Splatting (3DGS) is a 3D reconstruction technique that represents a scene as millions of volumetric gaussians, producing photorealistic results rendered in real time in a web viewer or in virtual reality.
In five stages: define the visual objective and the key areas, plan the capture route (light, access, overlap), capture photography and video on site with point cloud support where the scene needs it, process the 3DGS model checking continuity and navigation, and deliver it embedded in a navigable web viewer.
LIDAR point clouds and photogrammetry exist to measure accurately; Gaussian Splatting exists to show and explain a space at photorealistic quality. They combine well: LIDAR to measure, 3DGS to explain.
The model is explored directly in a browser, with no software to install and free movement through the scene, and can be shared by link with clients, public authorities or non-technical stakeholders anywhere in the world.
No. Traditional photogrammetry produces point clouds from imagery, from which DTMs, DSMs, orthophotos, meshes and metric point clouds are derived — accurate, but without hyper-realistic visual quality. Gaussian Splatting does not measure: it renders the scene in high definition with realistic lighting, for navigation and presentation. The two are combined to get the best of both.
A Gaussian splat is usually saved as a PLY file, or as SPZ when compressed. It needs a viewer that can render gaussians; a mesh or point cloud viewer will not display it properly. We deliver it already published in VisionHub, which opens PLY and SPZ in the browser and lets you share it with a link. For editing or cleaning up the splat there are tools such as SuperSplat.
Both start from photographs, but they produce different results. Photogrammetry computes measurable geometry (point cloud, mesh, orthophoto or DTM) and its visual quality is limited. 3DGS optimises millions of gaussians to reproduce how the scene looks from any viewpoint, including reflections and transparency, but it is not designed for measuring. One is for working with dimensions; the other is for showing the space.
Tell us what you need surveyed. We review your case with no obligation and send back a technical and commercial proposal matched to your requirements — in English, working to your project timetable.