Reconstruction API

Send photos from any camera, drone, or device. Get back accurate 3D models from the same engine that powers Polycam.

Book a demo Read API docs

Trusted by Half of the fortune 500

How it works

Book a demo Read API docs

Upload photos

Your product sends photos from any camera, drone, or device.

We build the model

Polycam's cloud engine does the work, no hardware on your side.

You get the result

The finished 3D model lands back in your system automatically.

Works with any capture source

Drones, phones, cameras, robots

Handles big jobs

Uploads built for very large photo sets

Fully automatic

Your system gets notified the moment a model is ready

Mobile app screen displaying a construction plan for Building 2 with a 3D structural model and progress percentages for finishing work at 28% and structural work at 81%. Overlaid is a code snippet showing JSON data about a reconstruction event marked complete, mode gaussian_splat, metadata with building, capture date, phase, and outputs including mesh and point cloud details.

What it does

Two outputs, one integration. Send the photos, we handle the rest, your product gets the model.

Book a demo Read API docs

Click and drag to experience

Photogrammetry

Photos in, accurate 3D models out. Objects, job sites, assets, evidence.

Click and drag to experience

Gaussian Splatting

Lifelike, explorable 3D scenes from photo sets. Built for viewers, product configurators, and digital twins.

What People Build

Screenshot of SkyBrowse software showing a 3D model aerial view of an NFL stadium in Cleveland, Ohio, with a code window displaying project reconstruction details including photo count and model outputs.

Drone and aerial platforms

Flight photos become 3D models automatically. No desktop processing, no GPU hardware, no per-seat software.

Digital insurance claim platform interface showing claim CLM-28473 for a single-family home with water damage, including a 3D model of a damaged room exposing walls and floors, alongside code output detailing claim metadata and reconstruction status.

Insurance and claims platforms

Adjusters capture on any phone; your claims system gets the 3D model back automatically. Documentation flows into the claim, not a camera roll.

Computer screen showing an industrial equipment management interface for a centrifugal pump labeled P-101 in the Boiler Room of Plant A, displaying a detailed 3D model of the pump with pipes and motor components on a concrete base, alongside a code snippet window titled 'bash' showing JSON data about photogrammetry reconstruction including event status, photo count, asset ID, rig ID, and schedule.

Robotics and inspection

Automated capture rigs collect photos on schedule. Polycam builds the models, results land in your asset system. Condition monitoring without a person in the loop.

User interface of construction site software showing a 3D model of an unfinished multi-story concrete building with scaffolding and construction equipment, sidebar menu with options like Overview, Models, Progress, Sheets, Issues, Team, and Settings, project progress at 68% labeled on track, and a black code window displaying JSON data about reconstruction completion, photogrammetry mode, photo count, project metadata, and mesh output details.

AEC software and integrators

Add capture-to-model to your own product. Site documentation, as-builts, and progress tracking powered by our engine, branded as yours.

Screenshot of a 3D floor plan model of an apartment unit with measurement labels on rooms, shown in a software interface named BuildAI with a sidebar and an AI Assistant chat displaying floor area info, and an overlaid terminal window showing a JSON output related to photogrammetry reconstruction data including event status, photo count, and scale details.

AI products and agents

Ground your AI on measured reality. A Polycam model is a measurement, so estimates, takeoffs, and answers built on it hold up. Your AI consumes the output; Polycam makes it.

User interface of a furniture catalog application showing 3D models generated from photo sets. The screen displays a beige sofa 3D model, photo thumbnails, and a sidebar with assets like sofa, chair, table, lamp, and plant with their photo counts and processing statuses. A code snippet overlay shows JSON data confirming completion of a 'gaussian_splat' reconstruction process with details such as photo count, batch information, and SKU.

3D content and simulation

Turn photo sets into game-ready models and Gaussian splats in bulk. From product catalogs to training environments.

Read API docs Book a demo

Security & Compliance

Read API docs

SOC 2 Type I & Type II

Icon of a key, a shield with a padlock symbol, a toggle switch, and two rectangles representing documents, symbolizing security and access control.

Access keys are protected, scoped to your workspace, and can be shut off anytime

Icon showing a document with an upload arrow, a padlock symbol for security, a clock for time, and a crossed-out key indicating restricted access.

Uploads use secure, expiring links; no standing credentials

Icon showing a web browser window with three listed items, each with a signal strength bar, alongside a globe symbol with a check mark indicating verified or secure global internet connection.

Public status page

Icon of a document with code brackets and a downward arrow symbol indicating code file download.

Downloadable API spec for engineering teams

Trust Center

Contact us

Book a demo Read API docs

The Reconstruction API is in limited availability on Enterprise.
You pay per model with prepaid credits. Your balance and full history are always visible.
Fair-use limits stated plainly, with a note that approved integrations get higher ones.

FAQs

What do I send?

Photos. Your product uploads a photo set through the API, we return the finished model. No preprocessing, no special file formats, just the images your capture source already produces.

What cameras and devices work?

Any camera that takes photos: phones, drones, DSLRs, robots, fixed-mount rigs. If it produces images, the API can reconstruct from it. There is no hardware requirement on your side.

How long does a model take?

It depends on the size of the photo set. Typical photogrammetry jobs finish in roughly 10 to 45 minutes, and Gaussian splats in roughly 30 to 55 minutes, with very large photo sets running closer to an hour. Actual times vary with image quality, overlap, and how well the photos align, so treat these as estimates rather than guarantees.

How is it priced?

Per model. You buy prepaid credits, each reconstruction draws down a set amount, and your balance and full usage history are always visible in your dashboard. Talk to sales for credit pricing at your volume.

Are there limits?

Yes, and they are stated plainly in the docs: caps on photos per job and concurrent jobs. These are fair-use limits, not a paywall, and approved integrations get higher ones as usage grows.

How do results get into my system?

Automatically. When a reconstruction completes, we send a webhook to your endpoint with the job metadata and secure, expiring download links for every output format. Your product picks up the model without anyone touching a dashboard.

Who owns the data?

You do. Your photos and the models generated from them remain yours, as described in our terms of service (https://poly.cam/legal/terms-of-service).

How do I get access?

Book a demo with the sales team. The Reconstruction API is in limited availability, so our team enables it on your workspace after a short scoping call.

‍