Skip to main content

RealityScan

TOOL
📖5 min read
Latest Version: 2.2
Making a Complete Model in RealityScan | Tutorial by RealityScan

Free From Epic

The application is a free download with no license fee. It is the successor to RealityCapture, which Epic rebranded in June 2025.

Masking Built In

Alpha masks exclude regions (sky, glass, moving people) from the reconstruction. That matters for site scenes, where the background is part of the problem.

GPU Pipeline

The reconstruction stages run on the GPU. NVIDIA cards are supported, and version 2.2 (2026) added full AMD support.

Introduction​

RealityScan is the free photogrammetry application from Epic Games. Since the 2.0 release it carries the name RealityScan; before that it was RealityCapture, and files, settings, and tutorials from the old name still describe the same pipeline. It runs the standard stages (camera alignment, dense cloud, mesh, texture) with a guided interface and a command-line mode for batch work.

Pick RealityScan when you want a free GUI pipeline with strong masking. Pick COLMAP when the run itself must be a reproducible command chain, Meshroom when you want an open-source node graph, and Metashape when you need a licensed product with georeferenced GIS outputs. The method is in the photogrammetry technique.

Installation​

  1. Download the desktop application from the official download page. Builds exist for Windows, macOS, and Linux.
  2. Install it through the package for your platform. No license key is required.
  3. Verify the version in the About dialog and record it. The current release is 2.2 as of June 2026.

Usage​

A reconstruction is a project file. You load images, optionally mask them, and run the pipeline stages.

Create a project and import images· project file

Start a new project and import the image set. What goes in: photos or extracted frames. What comes out: a project file that stores the image list and every stage setting. Name it after the case. If your photos carry GPS EXIF data, the application uses it for scene scale and position.

Apply masks· optional

Import alpha masks to exclude regions from the reconstruction: sky, water, glass, or people who moved between frames. Create the masks in any image editor with a transparent channel. Skip this step only when the scene has no problem regions.

Align the cameras· camera poses

Run the alignment stage. It estimates every camera position and the sparse cloud. Inspect the camera positions in the 3D view: a clean shell around the subject with no gaps means the coverage held.

Build the point cloud and mesh· cloud, mesh, texture

Run the dense cloud stage, then the mesh and texture stages. The dense cloud is the geometry you will measure against, so check it against known features before you accept the mesh.

Export· FBX, OBJ, USD, PLY

Export the textured mesh (FBX, OBJ, or USD) and the point cloud (PLY). What comes out: model files for Blender plus the project file and the stage settings for the case record.

In an Investigation​

Use the block below to document how you ran RealityScan on the evidence.

Documentation: how to document RealityScan

In Your Method Section​

Guidance: Describe the purpose, the tool version, and the input data.

  • Investigative Objective: To reconstruct the 3D geometry of the scene from a set of 2D images for spatial analysis.
  • Software Version: RealityScan [Version] (e.g., 2.2). Note that the tool was named RealityCapture before 2.0 (2025).
  • Input Data: [Image set name/ID], [N] images derived from [source footage/document], Source Hash: [First 6 digits].

In Your Decision Log​

Guidance: Record the specific procedural parameters and integrity checks.

  • Specific Settings: Masks applied (what was excluded and why), alignment and mesh quality settings, GPU used, EXIF GPS data present or stripped.
  • Assumptions: The scene was static during capture; masked regions contain no geometry you need.
  • Integrity Check: Analysis ran on a working copy, never the original evidence file.
  • Date Performed: [YYYY-MM-DD].

Verification & Mitigation​

Guidance: How did you verify accuracy and mitigate bias?

  • Verification: Inspected the camera shell in the 3D view. Compared the dense cloud against known feature dimensions.
  • Bias Mitigation: Masked moving objects instead of letting them register as ghosts. Kept the original EXIF data on a separate copy so the strip does not destroy the source.

Example Methodology Statement​

"Scene geometry was reconstructed with RealityScan 2.2 from [N] images of [source] (Hash: A1B2C3). Alpha masks excluded [regions] from the reconstruction. Camera alignment registered [N] of [N] images; the dense cloud was verified against a reference dimension of [X.X m] before meshing."

Into Blender​

RealityScan exports the formats Blender reads.

  • Mesh (FBX, OBJ, or USD): import with File > Import > FBX (built in), the bundled Wavefront OBJ add-on, or File > Import > Universal Scene Description (.usd).
  • Point clouds (PLY): import with File > Import > Stanford PLY (.ply), or use the Photogrammetry Importer add-on for georeferenced placement.

Three caveats before you model against the result:

  • Scale: if the photos carried GPS EXIF data, the scene has a real-world scale. Without it, the scale is arbitrary like any SfM pipeline: anchor it to a known dimension.
  • Check the orientation of the imported object and rotate it onto the Blender ground plane.
  • For georeferenced scenes: place the reconstruction relative to the site with ground control points or the QGIS baseline (custom CRS and origin point).

Limitations​

  • The algorithms are closed: a third party cannot rerun the same code, only the same settings. The project file is the record they get.
  • The product direction follows Epic's roadmap. The 2025 rebrand from RealityCapture and the closing of the legacy RealityCapture licensing (2026) show the tooling shifts. Pin the version in your method section.
  • The same SfM weaknesses apply: reflective, transparent, and featureless surfaces, and moving objects you did not mask, produce holes and false geometry.
  • Georeferencing depends on the EXIF GPS data or ground control points you supply.

Resources​

Sources​

  1. RealityScan, official site
  2. RealityScan 2.2 release post, current version and AMD GPU support
  3. RealityScan download page, installers for Windows, macOS, and Linux