Factory in 10 minutes 🏭

Plant in 10 minutes 🏭

Industrial Viar has had one very funny problem for a long time. Everyone dreamed about how an engineer would put on a helmet, go into the digital workshop, look at the machine, find a fault and fix everything, saving a lot of money. Sounds beautiful! Only before this, someone had to assemble this workshop in 3D with their own hands: walls, machines, stairs, passages, communications, etc. For weeks. Sometimes for months. With a budget that makes a normal production worker immediately want to return to Excel, an A3 printout and the phrase “let’s discuss it in words for now.”

And then Gaussian Splatting bursts into this boring pain 💣

At Hannover Messe they showed a case where, instead of manual modeling, an object can be captured in about 10 minutes, then the data can be automatically processed in 3-4 hours and a photorealistic XR environment can be obtained. Not an ideal engineering model, not a CAD with millimeter precision, not a “fire victim’s apartment”, as was the case with classic lidar scanning, but a “space” into which you can go, look around, which you can show to the customer, conduct training in it, or throw on a layer of hints.

And this seems to be more important than another helmet with a better screen.

The main stopper of industrial XR was not only in hardware, but also in content. Making the environment was expensive, time-consuming and tedious. And when the creation of an environment turns from “a separate project for the quarter” to “filmed in the morning, we watch it in the afternoon,” the economics of the whole story change. Logic becomes almost everyday.

There is a real object. It's quickly shot in 3D. Get a photorealistic XR scene. All the necessary information is added to it. And voila! 💁♂️ Industrial XR ceases to be a “look, we also have our own metaverse” attraction - it becomes a normal working environment. You can gather engineers from all over the planet in one space and discuss not an abstract drawing, but a specific workshop, right down to where a person will swear, hitting his head on a pipe. You can train employees not using a PDF with a dull “turn the valve” arrow, but inside a recognizable object.

However, there is some tar in this honey: such models are still mostly static. That is, the workshop can be shown. You can see the machine. The production site can be discussed. But showing work processes in dynamics is not possible. This is a new level of pain.

It's time to remember Gracia AI!

Dynamic video versions of Gaussian Splatting exist. Yes, this is not yet a mass industrial tool. But the vector has already been set: a quick scan of the space followed by the addition of live scenarios. At the same time, Gaussian Splats are not required to compete with other technologies. They can be mixed with 360 videos, regular 3D objects, CAD/BIM models, interactive tags, avatars, voice assistants and enterprise databases. The result is not “a new religion instead of old engineering,” but a working hybrid. CAD and BIM provide precision. Gaussian Splatting provides realistic visual context. XR enables shared presence and an understanding of scale. And now it looks like an adult market! Not a story about how hard workers will start working tomorrow in VR headsets, but a real case about how companies are starting to save on the most expensive part of the XR pipeline - creating an environment.

If earlier industrial VR was often sold through “wow, you can walk around a virtual factory,” now a more rigorous argument appears: it is possible to coordinate faster, train cheaper, sell more clearly and make fewer mistakes before the mistake is cemented with a certificate of completion.

When XR content can be collected not over months, but in one working day, it is no longer a toy for an innovation stand. This is a tool where the math finally starts to add up 🔟