From Real Object to 3D File: AI Tools and Scanners That Copy the Physical World

# From Real Object to 3D File: AI Tools and Scanners That Copy the Physical World

The idea of pointing a device at a physical object and pulling a usable digital model out of it sits at the intersection of augmented reality, machine learning, and precision laser scanning. Several widely circulated demonstrations suggest this workflow is becoming practical, but the strength of the evidence varies from case to case. This article walks through four of them, keeping claims tied to what the available sources actually show and flagging where they do not.

## Cyril Diagne’s AR App: Copy a Physical Object and Paste It as a 3D File into Photoshop

Designer and technologist Cyril Diagne is associated with a viral demonstration in which a phone-based augmented-reality tool appears to lift an object out of a real-world scene and drop it into a design canvas. Coverage of the demonstration describes an AR “copy and paste” gesture bridging the physical world and desktop software ([New World Notes](https://nwn.blogs.com/nwn/2020/05/cyril-diagne-ar-photoshop-capture-magic-leap.html)).

The specific claim that this app copies real-world physical objects and pastes them as *3D files directly into Adobe Photoshop* is not independently corroborated here. Diagne’s most widely shared “AR cut and paste” work is generally presented as capturing a 2D image cutout rather than a fully reconstructed 3D file, and the available source does not confirm a 3D-into-Photoshop pipeline. Readers should treat the “3D file into Photoshop” framing as an unverified description of the tool’s capability rather than an established fact. What can be said with more confidence is narrower: a Diagne-associated AR demonstration attracted attention for making the copy-from-reality-into-software gesture feel immediate.

## Software and AI Tools That Capture Real-World Objects as Digital 3D Models

Beyond any single demonstration, a broader category of tools aims to turn photographs or sensor data of a real object into a digital 3D asset. Coverage in this space describes AI-driven approaches that reconstruct an object’s shape and appearance from ordinary captured imagery ([80 Level](https://80.lv/articles/this-ai-algorithm-can-create-digital-copies-of-real-world-objects)).

In practice, tools that pursue this goal tend to fall into a few families: photogrammetry, which stitches many overlapping photos into a mesh; depth- or LiDAR-assisted capture, which uses sensor distance data; and newer learning-based reconstruction methods that infer geometry and surface detail from limited views. The available source points to an AI algorithm positioned in this last category. It does not, however, establish which method is universally best, how faithful the resulting models are across object types, or where the practical limits lie — those remain open questions rather than settled benchmarks.

## How an AI Algorithm Builds Digital 3D Copies of Real Objects

The claim that an AI algorithm can create digital 3D copies of real-world objects is one of the three assertions here that has not been independently cross-checked, so it should be read as a reported capability, not a confirmed one. The cited coverage presents such an algorithm and its output ([80 Level](https://80.lv/articles/this-ai-algorithm-can-create-digital-copies-of-real-world-objects)), but the source alone does not verify accuracy, reproducibility, or the conditions under which the results hold.

Described at a conceptual level, learning-based 3D reconstruction typically works by taking one or more images of an object, estimating the underlying geometry the images imply, and predicting surface color and texture to produce a model that can be viewed from angles the camera never directly captured. Where traditional photogrammetry needs dense, overlapping coverage, learning-based systems are often promoted for their ability to fill gaps from fewer inputs. Sources here do not confirm how well that gap-filling performs on complex materials, reflective surfaces, or fine detail, so the quality of any “digital copy” produced this way should be understood as claimed rather than demonstrated.

## Scanning at Scale: 3D Capturing the Interior of the Great Pyramid at Giza with the Leica BLK 360 and Matterport Pro 2

At the largest end of the spectrum sits the scanning of entire architectural interiors. An interactive walkthrough of the interior of the Great Pyramid at Giza has been published as a navigable 3D experience ([Mused](https://giza.mused.org/en/guided/266/inside-the-great-pyramid)), which demonstrates that a detailed digital capture of the pyramid’s interior exists and can be explored.

The more specific claim — that this interior was 3D scanned using a Leica BLK 360 together with a Matterport Pro 2 scanner — is not independently corroborated by the available source. The cited walkthrough confirms the *existence* of a 3D interior capture, but it does not, on its own, verify the particular hardware used to produce it. Both devices named are real, established tools in the reality-capture field: the Leica BLK 360 is a compact terrestrial laser scanner, and the Matterport Pro 2 is a camera system built for spatial capture, and either or both would be plausible choices for such a project. Plausibility, however, is not confirmation. Until the specific equipment is documented in a reliable source, the attribution to the BLK 360 and Matterport Pro 2 should be treated as an open, unverified detail rather than a settled fact.

—

Across all four cases, the through-line is the same: the ambition to copy the physical world into an editable digital form is well underway and, in the case of the Great Pyramid interior, has demonstrably produced an explorable result. The finer claims — a 3D-into-Photoshop pipeline, the fidelity of AI-generated object copies, and the exact scanners used at Giza — remain reported rather than confirmed, and are best held at arm’s length until stronger corroboration is available.

For another view of the 3D tooling ecosystem, see Anthropic Joins the Blender Development Fund as a Corporate Patron — Then Shifts to a One-Time Donation.