EN TR
Spatial Computing Research
Spatial Capture02 - ROOMField Test2023-2025

Spatial Computing

What if clients could walk through buildings before we build them?

Scale 02 Room
Capture LiDAR iPhone Pro
Accuracy ±2cm field-measured
Tools Unity Quest 3
Type Field Test
Updated 2026-07
01

Every architect knows this conversation: 'I thought this room would feel bigger.'

It happens after the walls are up. After the money is spent. The client approved floor plans they didn't really understand. Now they're disappointed. Change orders follow. Trust breaks.

The problem isn't the design. It's the communication medium. Floor plans are abstractions. Even 3D renders on a screen don't convey scale. You can't feel a room from a picture.

So we field-test a different workflow. LiDAR scans existing conditions in about 30 minutes. VR lets a client walk through the proposal at 1:1 scale. AR overlays new walls on the existing space. This note reports what we've measured ourselves (scan times, accuracies) and is explicit about what we haven't: a controlled study of how much miscommunication this actually removes.

High-resolution LiDAR point cloud

30 Minutes to a Working Model: A typical scan captures on the order of a million points and replaces a multi-day manual survey at centimeter accuracy.

02

Theoretical Framework

01

LiDAR Capture

30-minute scans for 60m² spaces on an iPhone Pro. Replaces multi-day manual surveys at accuracy that's honest about its limits.

02

VR Review

1:1 scale walkthroughs. Quest 3 for accessibility, Vision Pro for quality. Clients feel the space before it exists.

03

AR Verification

On-site overlay of the digital model. Contractors see clash zones before they become problems.

04

Shared Understanding

Clients approve what they've walked through. We haven't run a controlled change-order study; the literature and our field use both point the same direction.

03

Research Process

01

Scan with LiDAR

iPhone Pro captures a room in 15-25 minutes at roughly ±2cm accuracy

02

Build VR Environment

Point cloud to Unity via Rhino.Inside, a few hours for a low-poly spatial model

03

Walkthrough Session

1:1 review with annotations and furniture testing, 45-60 minutes

04

AR Site Verification

Overlay model on construction to catch discrepancies early

04

Research Phases

01

Technology Benchmarking

Comparing capture options at residential scale (iPhone LiDAR in hand against pro-scanner specs) for accuracy versus speed.

02

VR Pipeline

Point cloud to Unity workflow with Rhino.Inside. Optimized for spatial accuracy, not photorealism.

03

Field Trials

Capture-review loops run on our own spaces and concept studies. A controlled A/B with real clients is future work, and we say so.

04

SpaceCraft App

iOS app integrating capture and review in one tool; in development, rebuilt scan-first for v2.

05

Key Metrics

30 min
Scan Time
For a 60m² space
±2cm
Accuracy
iPhone LiDAR, field-measured
~1M
Points per Scan
Typical residential capture
1:1
VR Scale
True-scale walkthroughs
06

Key Thinkers

01

Jaron Lanier

VR Pioneer, 1980s

Lanier coined 'virtual reality' and built the first commercial systems. Our client-facing VR inherits his vision of experiential computing.

02

Apple Vision Pro Team

Spatial Computing, 2024

Apple made 'spatial computing' a mainstream term. Our RealityKit integration follows the architectural decisions they set.

03

Greg Lynn

Digital Design Pioneer

Lynn pioneered computational design in architecture. Our parametric-to-VR pipeline walks a trail he opened.

04

James Turrell

Light and Space Artist

Turrell's installations explore how humans perceive space. We prioritize perceptual accuracy over photorealism.

07

Where This Lives in Our Products

SpaceCraft: this research, shipped

In development

SpaceCraft is the capture side of this workflow made into a product: iPhone LiDAR scanning, rebuilt scan-first for v2. Every number on this page about scan speed and accuracy comes from building and using it.

SpaceCraft →

Scanned context for AI visualization

Exploration

A scan gives AI visualization true geometry to hang on. Feeding scanned rooms into Archly's workflows, so renders start from the space as it is, is an integration we're exploring.

Archly →

Comparative Analysis

LiDAR Scanning

30 Minutes to Model

iPhone Pro or professional scanner. Centimeter accuracy. Replaces tape measures and laser distance finders.

FastAccuratePortable

VR Immersion

Walk Through Before Build

1:1 scale walkthrough on Quest 3 or Vision Pro. Clients experience space, not just see it.

Experiential1:1 ScaleClient-Facing

AR Verification

On-Site Overlay

Digital model projected onto the construction site. Catches discrepancies before finishes go in.

Real-TimeOn-SiteVerification

Digital Twin

Living Document

Continuously updated model from scan + design + feedback. Single source of truth throughout a project.

ContinuousBIM IntegrationComplete
05

Optimization Results

100% 75% 50% 25% 0%
48%
28%
8%
2D Only
Screen 3D
VR Walkthrough

How often does 'I thought it would feel different' happen, by review medium?

Scenario model: expectation ranges from literature and our field use, not a controlled study

08

Key Findings

01

LiDAR saves days. A tape-and-laser survey takes 1-2 days; our scans of the same kind of space take under an hour, processing included.

Field-measured
02

Scale is the message. People read plans politely and read VR viscerally. The 'I thought it would feel bigger' conversation happens before walls exist, not after.

1:1 changes the talk
03

Accuracy is honest at ±2cm. Good enough for layout decisions, not for millwork. Knowing which decision is which is the actual skill.

±2cm
04

The literature points to fewer change orders with immersive review. We believe it, and we haven't measured it ourselves yet. That study is on the roadmap.

Plausible, unmeasured
09

Honest Limitations

Data Dependency

Our field runs are our own spaces and concept studies, not a controlled study with real clients.

Data Dependency

We haven't measured the change-order effect ourselves. The big reduction figures you read elsewhere are from the literature, not from us.

Computational Cost

Low-poly only. We test spatial VR, not photorealistic. Does visual fidelity matter for decisions? We don't know.

Behavioral Assumption

Capture quality degrades on glass, mirrors, and dark surfaces: the classic LiDAR blind spots.

10

Conclusion

A scan takes half an hour where a survey took days. A client who has walked a design at 1:1 knows what they're approving. That much we've seen ourselves, repeatedly. The rest (exactly how much money immersive review saves) is the measurement we still owe, and we'd rather owe it than invent it.

Limitations

  • No controlled client study yet
  • Spatial VR only, not photoreal

Future Directions

  • Controlled review-medium comparison
  • Photorealistic VR versus spatial VR