Gaussian Splatting in Luma AI – Moto Guzzi V9 Bobber
- 5 hours ago
- 2 min read
In recent years, there have been major advances in 3D visualization and scanning technology. What once required expensive equipment to digitize objects can now be done with nothing more than a smartphone.
To put that to the test, I created a Gaussian Splat of my Moto Guzzi V9 Bobber using Luma AI, and the result was surprisingly good.
What is Gaussian Splatting?
Gaussian Splatting is a relatively new technique within 3D visualization. Instead of relying on a traditional polygon-based model, as used in tools like Maya or Cinema 4D, this method works with point clouds, consisting of thousands of tiny points that carry color and depth information.
Luma AI converts a series of photos into such a point cloud and then calculates how these points come together to form a 3D image.
The result is a smooth, highly realistic representation of the scanned object, complete with light, shadow, and material reflection.
The test: Moto Guzzi V9 Bobber
For this first test, I chose an object with plenty of detail as well as reflective surfaces: my Moto Guzzi V9 Bobber.
I captured a series of photos from different angles, after which Luma AI automatically generated a 3D Gaussian Splat.
It was impressive to see how well the app managed to reconstruct shape, color, and depth. Even small details like the rims, tank reflections, and engine components came through with surprising accuracy.
Endless possibilities
What excites me most about this technique is its accessibility.
Anyone with a modern smartphone can now capture 3D objects at a level of quality that would have been unthinkable just a few years ago.
For designers, visualisers, and marketers, this opens up a wide range of new possibilities:
Product visualisations: scan a physical product and use it directly in an animation or AR presentation.
Client cases: create interactive 3D representations of completed projects.
Archiving: digitally preserve unique objects while retaining their texture and form.
Gaussian Splatting also introduces a new kind of realistic, non-destructive 3D representation that can be processed extremely quickly.
Unlike traditional photogrammetry, the results are visible almost immediately and require far less heavy post-processing.
