Why we still scan with Matterport in 2026

The 3D capture space exploded since 2023. Polycam and Luma made photogrammetry phone-pocketable. Gaussian splats produced renders that read more like memory than geometry. Apple's RoomScan landed in iOS. The whole field was supposed to make Matterport feel like a fax machine.

It didn't. Not for the work we do. And the reason is unromantic.

Determinism beats wow

A virtual tour for a Porsche dealer doesn't need to wow the engineer who scanned it. It needs to load on a Samsung from 2021, embed inside an iframe nobody chose, indexed by Google, and read the same in three years as it does today.

Matterport's stack does that. The output is a known quantity. The hosting is a known quantity. The embed is a known quantity. The model is owned, indexed, citable, and won't disappear when a startup pivots.

The competitive workflow

Yes, photogrammetry can produce higher-resolution geometry. Yes, splats can render lighting that Matterport flatly cannot. We use both. But not for the same brief.

For a museum hall that needs to be archive-grade and citable, Matterport wins. For a single hero object that needs to be viewed at film-set fidelity, splats win. For a marketing render of a space that doesn't exist yet, photogrammetry of mood + Unreal beats both.

The mistake is choosing a tool by what's new instead of what the brief actually needs.

What we're watching

Matterport's Genesis camera is a step-change in geometry quality. Niantic's Scaniverse is making smartphone splats actually usable. The first product to combine deterministic hosting + splat-quality rendering will be a quiet revolution. We'll switch the day someone ships that.

Until then: the unromantic answer is that boring tools that ship the brief beat exciting tools that don't. We're paid to ship the brief.

FAQ

Why do you still use Matterport in 2026 when newer tools like Gaussian splats and phone photogrammetry exist?

Because for our work, determinism beats wow. Matterport's output, hosting, and embed are known quantities: the model is owned, indexed, citable, and won't disappear when a startup pivots, which matters for a tour that must load on an older phone, embed in an iframe, be indexed by Google, and read the same in three years.

Do you only use Matterport, or do you work with photogrammetry and Gaussian splats too?

We use all three, chosen per brief. An archive-grade, citable museum hall suits Matterport; a single hero object at film-set fidelity suits splats; and a marketing render of a space that doesn't exist yet is better served by photogrammetry of mood combined with Unreal.

How do you decide which 3D capture tool to use for a project?

We choose by what the brief needs, not by what is newest. Choosing a tool because it is new rather than fit for the job is, in our view, the core mistake.

What new developments in 3D capture are you watching?

We're watching Matterport's Genesis camera for a step-change in geometry and Niantic's Scaniverse for usable smartphone splats. The first product to combine deterministic hosting with splat-quality rendering would be a quiet revolution, and we'd switch the day someone ships it.