Turn the real world into a model you can get answers from.

We reconstruct real places from ordinary video, imagery, satellite data and LiDAR, then answer practical questions on top of them.

See how it works
Source

Loading the scan · 0%

Capture

Start with the evidence you already have.

This office was captured in one unstructured walk with an ordinary phone. No LiDAR, no capture procedure, no special hardware. The amber line is the path the camera took, recovered from the video itself.

Phone videoDrone videoPhotographsArchival footageLiDAR & point cloudsSatellite imageryExisting scansAn address
6:22of phone video
1,900camera poses
0lidar sensors

Reconstruct

Every frame becomes geometry.

We work out where every pixel sits in space and fuse the frames into one reconstruction, to scale. The room you walked through is now something you can view, and measure, from any angle.

For exteriors we gather the available satellite imagery, drone data and public LiDAR for you.

10.6Mpoints reconstructed
Metricscale

Understand

It knows what it is looking at.

Every point carries a label, so you ask for objects by name instead of tracing them by hand. Pick one and the rest of the room steps back.

Structure

Clean geometry. No Manhattan assumptions.

A point cloud is only the beginning. Our neural geometry system recovers the structure as boundary geometry (faces, edges and vertices), the representation CAD, estimating and simulation tools already consume.

Walls are solved where they are and heights per surface, so angled, sloped and curved geometry survives. No hand-tuned fitting rules per scene.

78faces
46planes
11rooms
Shared edgeJunction

Measure

Measure it. Then price it.

The model is to scale, so lengths, areas and quantities come straight off it. Pick a room on the plan or in the model.

Quantities flow into itemized repair estimates with location-specific pricing, using BNI Costbooks as the cost data source.

Ask

One model. Any number of questions.

Build the model once and every workflow that follows runs on it. No manual remodelling between questions.

  1. 01What is there? Walls, openings, fixtures and damage, labelled.
  2. 02How big is it? Dimensions, areas and quantities, to scale.
  3. 03What is damaged? Damaged regions found and measured on the model.
  4. 04What will it cost? An itemized, location-priced estimate.
  5. 05What do I hand over? CAD-ready geometry and a photorealistic walkthrough.

The problem

Every decision about a place begins with a measurement.

Every claim, repair, recovery grant, resiliency study and risk decision begins with a site measurement, but measuring well still involves trade-offs.

An office floor cut away in isometric view, with a camera's path through its rooms dissolving into a point cloud

Survey-grade scanning

Delivers The most accurate results.

But requires

  • Expensive equipment
  • A trained crew
  • A site visit

Phone capture tools

Delivers Rooms captured in minutes.

But often

  • Require guided capture
  • Assume straight, orthogonal walls
  • Produce sparse or simplified reconstructions

The footage crews actually produce

These tools struggle most with it.

What it looks like

  • An unstructured walk-through
  • A drone pass
  • A lower-quality video recorded years ago

It either goes to a modeler or goes unused, leaving decisions to be made from photos and tape measurements.

The office reconstructed above is that kind of footage: one unstructured walk with an ordinary phone, no capture protocol, specialized hardware or training.

Progress

From ordinary footage to clean geometry and a priced estimate.

Geometry from noise

The neural reconstruction system produces clean boundary geometry from point clouds of different origins without scene-specific fitting rules.

Field to estimate

A working iOS prototype converts field photographs and video into an itemized, location-priced repair estimate.

Insurance and risk

In August 2026, Justmagic signed a cooperation agreement with Athenium, an insurance and risk analytics company.

Industries

Built for industries where understanding a place has economic consequences.

Property measurement, construction estimating, insurance, disaster management, infrastructure and defense already pay separately for capture, measurement, simulation and cost analysis. Our core technology can connect those functions.

A reconstructed house with a damaged area marked on its roof and the repair estimate it produces

Property claims

Roughly five million U.S. homeowners claims are filed each year, averaging about $17,000 each. Every one starts with what exists, what was damaged, how much of it there is and what it costs to repair. We turn the evidence collected during the claim into that structured understanding.

$85.7Bspent on P&C claims adjustment in 2024
A reconstructed neighbourhood, every structure inventoried and shaded by how badly it was damaged

Disaster recovery

Large-scale recovery programs need inspection, damage assessment, measurement and technical analysis across thousands of properties, with site inspection and field services in scope. We reconstruct sites from evidence that already exists and apply the same measurement and analysis stack across all of them.

$2.3BFEMA Public Assistance technical-assistance ceiling
A reconstructed house standing in flood water, with wind streamlines passing over its roof

Risk and engineering

Catastrophe models begin at the hazard footprint. We extend the analysis down to the site itself: flood depth around a particular structure, wind load on a particular roof, blast pressure on a particular wall.

$2BMoody's acquisition of RMS, the scale of physical-risk modeling

Paid for separately today

  1. Capture
  2. Measurement
  3. Simulation
  4. Cost analysis

Potential customers and integration partners

  • Established platforms
  • Engineering firms
  • Insurers
  • National, state/provincial and municipal governments worldwide

Applications

The same model, taken outside.

The office above shows the technology. Each application runs it on a different kind of place, from the footage that place already has.

A house reconstructed as a point cloud from above, its roof solved into coloured planes

Exteriors and roofs

Measure a roof without anyone climbing it.

We reconstruct the house from ordinary imagery, find the roof and solve every facet as a plane, ready for the estimate.

PitchAreaRidgesValleysEavesRakes
See the exterior walkthrough
Downtown San Diego reconstructed in 3D mid-storm, flooded buildings tinted yellow to red by depth

Floods

See which buildings take water, and when.

We reconstruct a city from the footage that already exists and play the storm forward over it, for rescue triage and recovery cost estimates.

Depth per buildingTime of arrivalMaximum inundationSensor updates
See the flood model

Origin

Born from disaster recovery.

san juan, puerto rico · 18.47° n · 66.11° w · est. 2023

Justmagic LLC was founded in San Juan, Puerto Rico in 2023, out of the recovery from Hurricane Maria — where measurement, recovery and risk were not abstractions.

We create a version of the real world on a computer, so the decisions that depend on it can be made faster and more precisely.

CEO

Vladimir Santalov

Spatial statistics.

Delivered $16 billion in recovery funds for Puerto Rico after Hurricane Maria through stratified statistical sampling and LiDAR feature extraction; modeled Hurricane Harvey flooding for Houston.

CTO

Breno Jesus

Computational mathematician and AI researcher specializing in robotic vision.

Built the neural network that recovers clean geometry from noisy scans.

Answers

Frequently asked questions

01What can I give you to work from?

Phone video, drone video, photographs, archival footage, LiDAR or point clouds, satellite imagery, existing scans in most common formats — or simply an address. Describe the area you care about in plain language.

02Do I need special hardware or a capture procedure?

No. There is no prescribed capture sequence, specialized camera or training workflow. Slower movement and more overlap give a denser reconstruction, but a phone without LiDAR is enough to start.

03What do I get back?

One reusable model with three views of the place: a structured scene, with walls, windows, doors, roofs, floors and damage identified automatically; CAD-ready geometry, with point clouds converted into clean faces, edges and vertices for CAD and simulation; and a visual model, with detailed interiors and photorealistic walkthroughs.

04Do I need to set up the pipeline myself?

No. Our AI agents assemble the right pipeline from our proprietary reconstruction and simulation tools, hiding the complexity from the user. There are no hand-tuned fitting rules or scene-specific adjustments.

05What does "structured geometry" mean?

Walls, roofs, floors, openings and damaged regions recovered as clean boundary geometry — faces, edges and vertices — rather than a cloud of sampled points. It is the representation CAD, estimating and simulation systems already consume.

06Is the model to scale?

Yes. The model carries metric scale, so areas, spans and volumes are measured off it directly rather than inferred.

07What is a Manhattan assumption, and why does it matter?

It is the shortcut of assuming every wall meets at ninety degrees and every floor and ceiling is flat and level. It makes reconstruction much easier — and it makes the quantities wrong the moment a building has an angled wall, a sloped ceiling, or a curve. We do not make it.

08Does it work on exteriors?

Yes. For exteriors we assemble drone and phone video, photographs, satellite imagery and aerial LiDAR — or start from an address — and solve the roof facet by facet: pitch, area, ridges, valleys, eaves and rakes. See the exterior walkthrough.

09What kinds of analysis can run on the model?

Measurements and quantities, damage identification, itemized locally priced repair estimates, CAD-ready geometry and photorealistic walkthroughs. One capture, one model, any number of questions.

10How are estimates priced?

With location-specific pricing. We use BNI Costbooks as our primary cost data source and apply regional cost factors for the site's location.

Get started

The real world already contains the answers.

We make them computable. Bring us a video, a scan, an archive — or an address.