Coming soon · English first

There’s a world
in every word.

Explore a scene. Follow a detail.
Find the words for real life.

Mavilo’s curious poodle tilts its head over an open sage-green book.
A sunny café opens onto a leafy square. A barista hands a coffee to a customer, and an ivory latte cup rests on the oak counter. Look closer
One scene. A world of things to discover.
A sunny café. An ivory latte cup sits on the foreground oak counter, while a barista hands a takeaway coffee to a customer. Scene · At the café

Small detail.
New connection.

Layer 0

Scene · Choose an entry point

Start with a familiar scene.

You are ordering at a café. An object or a phrase can open a different path into the same situation.

Why it fits this scene

“Coffee” opens the drink and its details. “To go” opens the way you ask to take it with you.

Choose a starting point. Each opens a connected explanation.

An illustrated method preview. Follow the café into a drink, a cup and its handle — or explore the conversation. Five prepared views, connected explanations and a trail you can return to.

Curiosity loves company.

Choose a companion. Find the personality that fits your way of exploring.

Poodle

The Guide

Mavilo’s main mascot

A curious poodle companion who helps you find what to explore next.

  • Curious
  • Practical
  • Patient

Companion concept · In testing

Choose a personality to preview. Companion concepts are being tested.

Looking ahead · 3D under explorationA spatial future. The same curiosity.

A photographed place could become another way into the same Drill-Down Explainer. This is our research direction; the interactive preview above demonstrates our 2D starting point.

An original café concept illustrated as a spatial cutaway, with its outer edges dissolving into soft particles.
Concept illustration · Not a live reconstruction or a 3DGUT render.

01 / The starting material

A place, captured from several views.

Photographs and calibrated camera views provide the input for reconstructing a spatial scene.

Mavilo’s 2D MVP starts with a generated illustration. This proposed 3D path starts with photographs of a place.

Where NVIDIA’s 3DGUT fits

More faithful cameras.
More believable scenes.

3D Gaussian Splatting represents a photographed scene with soft 3D particles. 3DGUT improves how they project through complex camera models, helping preserve the scene as the view changes.

  • Distorted lenses, accounted for.

    It supports nonlinear camera models, including fisheye lenses, rather than relying on an ideal pinhole camera.

  • Camera motion, accounted for.

    It can model rolling shutter: different parts of a frame captured at different times.

  • An efficient rendering path.

    It retains Gaussian rasterization efficiency. Mavilo’s frame rates and supported devices still need validation.

What about reflections, refractions and the Unscented Transform?

3DGUT means 3D Gaussian Unscented Transform. Representative “sigma points” pass through the camera model to estimate each particle’s screen footprint.

A 3D particle’s sigma pointsThe camera modelA screen footprint

Reflections and refractions use hybrid rendering: the main view is splatted, while secondary rays are traced with 3DGRT. These effects do not come from splatting alone.

Language labels, object selection and explanations would be Mavilo’s product layer. The rendering research does not supply them.

3D remains a research direction for Mavilo. We have not validated real-time performance, device compatibility or learning outcomes. The illustration above demonstrates the idea.

Mavilo’s poodle peeks curiously over a large sage-green book, with both paws holding it open.

A little curiosity
goes a long way.

Coming soon · English first

What would you like to learn?

Preview only. Your email is not submitted or saved.

A place at the beginning. A recognition badge for the first 100 trial participants. This preview does not reserve a place.