Anyone can make 3D now. So why don’t job sites use it? We asked the maker of a Gaussian splatting viewer

Take a video on your phone, and as soon as the next day you can walk through a 3D copy of that street that looks like a photograph. In three years, that went from research demo to everyday tool.

The technique is called 3D Gaussian Splatting, or 3DGS. Researchers at Inria, the French research institute, and their partners published it in the summer of 2023. In September 2026, it was formally added to glTF, the common file format that lets 3D data move between different software, a bit like JPEG does for photos. From paper to shared standard took just three years.

At DATA WORLD we have been following PLATEAU, Japan’s national program of 3D city models. Making 3D data keeps getting easier. Using it as an everyday work tool, in site meetings or when explaining a project to local residents, is another story.

Japan’s infrastructure ministry (MLIT) surveyed building teams in fiscal 2024. Even among teams that already design in BIM, the 3D modeling standard for buildings, 38.6% had never shared their 3D data with other departments or outside firms. Another 45.7% said they couldn’t, because their subcontractors don’t use BIM. That is despite 82.7% saying 3D makes communication better. Easy to make, hard to hand over: that is the wall.

So the tools to make 3D are here. What is still missing?

One company is answering that from the user’s side: World Area Networks (WAN), based in Hiroshima.

We sat down with WAN’s CEO, Ayumu Mitani.

The short answer: what’s missing isn’t better capture. It’s a tool that turns 3D into a conversation where everyone knows exactly which spot you mean.

WHO IS

Ayumu MitaniAyumu Mitani

CEO, World Area Networks

“Once you’ve made the data, how do you actually use it? The software hasn’t caught up yet.”── Ayumu Mitani, in this interview

1. 2010From music school to founder at 25After music college, the company he joined went bankrupt within a year. He discovered CG and VR and founded WAN in Hiroshima
2. 2018–Into constructionStarted building 3D simulators for construction machinery, leading to the excavator simulator Metatore and the 3DGS viewer Machica
3. NOWToward physical AIMetatore+, which uses AI to pass on veteran operators’ skills, was selected for a national R&D program. Physical-AI research formally launched in October 2026

He started out making things to look at, with video and CG. Then came virtual spaces where people practice operating machinery, and now 3D copies of real job sites.

It stands for 3D Gaussian Splatting. It’s a new kind of 3D graphics.

You’ve probably heard of digital twins: copying a real place into virtual space so you can check things or try things there. A lot of companies are now working on how to reproduce real places in virtual space, and 3DGS is the technique everyone is watching.

You take photos or video from lots of angles. From those, it builds the scene out of millions of tiny oval particles, each with its own color and transparency. The particles don’t describe the shape of an object. They describe how it looks. That’s why reflections in glass and the fine detail of leaves come out like a photo, and the scene still looks natural from any angle.

A point cloud describes shape with dots. From a distance, you see something like a street. Up close, you see all the gaps between the dots, and it’s hard to tell what’s what.

What point clouds do well is measurement. You can measure from this point to that one without going back to the site, even things you forgot to measure on the day.

With 3DGS, you get a 3D model that looks almost exactly like the real place, and you can have it in as little as a day. Some 3DGS data carries point-cloud data underneath, and 3DGS itself has real-world scale, so you can still measure distances, though the accuracy is a bit different from a point cloud. Swap 3DGS in for point-cloud work, and everything becomes much easier to understand at a glance.

Mostly you just walk around with a handheld scanner. You can even use iPhone video. A rough pass already works, and if you shoot the parts you care about up close, they come out beautifully. You can read the text on a shop flyer.

If you need it for surveying, you can place markers at a few points and record where they are. But even a casual capture looks very good. People with no experience can do it.

More and more people are making 3DGS, with handheld scanners, drones, iPhone video. Everyone is still experimenting with how to get clean data.

But what happens after you make it? That part is just missing. You get a beautiful 3D model, and there’s no tool that lets people on site open it, measure it and talk about it. The software for using it hasn’t caught up with the technology for making it. So we thought: fine, we’ll build the user side ourselves. That’s how a viewer made just for 3DGS started.

The name comes from “machi” — Japanese for town — and “visualize.” Making the town visible.

WAN has supplied a few systems to construction and civil engineering, like our machinery simulators. Machica is a tool for putting 3DGS digital twins to work on those same sites.

Everything happens in the browser. You issue a URL, and whoever receives it — a client, residents at a briefing — can view it for free. A bit like YouTube. An admin uploads, and anyone can watch.

For security, you can also share it privately or with a password. Show it to just the people who need it, or to everyone.

The 3DGS data carries survey coordinates — where in Japan it is. BIM design data is usually built on survey coordinates too, so when you import it, it snaps right into place. It fits really, really precisely.

If there are no coordinates, you can place it by hand, the way you would in CG software.

Honestly, we came in with no construction experience, so at first we built it on guesswork. We handed it out to lots of customers, and right now it’s free as a beta.

Then the people using it tell us, “This would be easier,” or “We need this feature.” Bit by bit, we’ve come to understand the industry, and we add whatever makes sense from there.

We get some really passionate feedback, and we’re grateful for it. The people interested in 3DGS right now tend to be early adopters, people who want to try new things.

Measurement, for one. To us, you just measure from here to there. In practice, what people actually want to measure is slightly off from where you can click, so they end up eyeballing it. Users have taught us, “Measure it this way and it’s accurate.”

Another is driving simulation. People building roads or railways want to see what it looks like to drive along it once it’s finished. Feedback like that makes us go, “Oh, people use it like that,” and we keep adding features.

I don’t think many are using it fully in production yet. They’re testing it on real jobs, and only a handful of companies at that. I expect that to spread quickly.

That said, we exhibited at CSPI-EXPO, a construction and surveying trade show, in June 2026, and “3DGS” was everywhere. It really drove home how much attention it’s getting.

Machica works with static 3DGS today. But 3DGS that captures time is already appearing, with people moving inside the scene. You play back a past moment inside a digital twin. Video that has become three-dimensional. Like something out of Star Wars.

When that arrives, how will we use it? I honestly can’t see that yet, but I’m sure the uses will come.

I come from video production, so I want to use it for video. Being able to change the camera angle afterward is such a relief. “Oh no, I forgot to get that shot” — with 3DGS, you can just shoot it later. Once you have the data, you can film as much as you like.

Reflections in glass hold up even as you move the camera. If you end up shooting on a rainy day, you might be able to swap out just the sky.

Real estate, tourism, the metaverse, games — it fits all of them.

You can make a simple 3DGS with an iPhone app now, so you could even 3D-print one. And now that non-engineers can build things with AI, instead of modeling a game’s background from scratch, you could capture a real town as data. People have talked about traveling in VR for years. Now you’d walk through a genuine 3D copy of the place, wherever you want to go.

Right now we’re very close to construction and civil engineering. Our first goal is to get those customers using it.

First, make it a tool that’s genuinely used where it’s needed most. From there, I think we’ll see other directions — real estate, tourism, urban planning — and branch out step by step. When you need to talk about a real place, you open Machica first. That’s the everyday tool I want it to be.

Leaving comments and documents right on the 3D space.

Construction sites have mountains of documents. “Which folder was that file in again?” If they live on the space itself, you can pull them up easily and share them with clients. And when you’re trading comments and can’t tell which spot a fix refers to, on the 3D scene it’s instantly clear: “Oh, here.”

Think of it as a spatial Slack, a communication tool tied to spatial data. If people use it that way, site discussions get faster and more accurate.

Shoot it on a phone, have it in a day, share it with anyone by URL. If that keeps going, here is the kind of job site it points to:

• Sites get captured every day
  From a distant office, you can walk today’s site in 3D. Missed measurements and missed shots can be recovered later.
• Meetings happen on the 3D model
  Clients, residents and crews point at the same spot in the same scene. People who can’t read drawings can join the conversation.
• The 3D record becomes the base for the next job
  Construction records carry over into inspection, maintenance and town planning.

WAN is already looking further ahead. In July 2026, its research to add AI that evaluates veteran operators’ skills to the excavator simulator Metatore was selected for a national R&D program. 3D copies of real places may become not just meeting rooms for people but training grounds where AI learns site work. We follow that thread in the next part of this series, a day with Mitani.

How to read this, depending on who you are

• If you run job sites: Start by capturing. Phone video is enough to make 3D now. The real value of 3DGS isn’t how pretty it looks. It’s how much faster it makes decisions with everyone involved
• If you’re a student or creator: Video, CG and game skills are becoming the skills for recording real places and building work tools around them. There are plenty of uses only a video person would spot
• If you’re an engineer or researcher: How do you bridge 3DGS’s visual realism and survey-grade accuracy? How do you log when data was captured and what was edited? With a shared format now in place, these are open problems worth working on

Beyond our own products, WAN also takes on custom production and development. Projection mapping, machinery simulators, Machica — at the root, it’s all copying real space into virtual space. So if you ever wonder, “Could this be done in virtual space?” or you have a problem to solve, just reach out. We have a lot of tools in the box, so we can usually find an answer.

In three years, the tools to capture and build 3DGS have largely arrived, and it now has a shared file format. Even so, by Mitani’s count, only a handful of companies use it fully at work.

What’s missing isn’t camera skill. It’s getting the 3D into conversations with clients, residents and crews. That’s exactly what Machica’s comments-on-the-model feature targets. As we see it, the value of 3D lies less in how good it looks than in making it instantly clear which spot everyone means.

“We had no industry experience, so we started with guesswork,” Mitani told us frankly. From there, WAN offers the beta for free and lets site workers teach it things as basic as how to measure. That’s how features no developer would have thought of, like driving simulation, appear.

Tools built only by experts tend to work only for experts. A viewer made by a video person becomes something residents who can’t read drawings understand. There’s a real lesson in that pairing for construction tech.

With 3DGS now part of a shared 3D format, splat data is no longer locked inside one app. The next question is whether the comments and documents attached to a space can travel with it to other tools.

Can viewers like Machica bring the record of site conversations into that shared ecosystem? That may decide whether 3DGS truly becomes a work tool. We’ll keep watching.