RealityScan Releases V2.1.1 with COLMAP Updates

Michael Rubloff

Michael Rubloff

Email
Copy Link
Twitter
Linkedin
Reddit
Whatsapp

Epic Games has followed up with version 2.1.1, a version that will hopefully significantly benefit people who are using RealityScan for their SfM before doing reconstruction. While 2.1 introduced automation, remote execution, and expanded LiDAR support, this release narrows in on the details that determine whether datasets move cleanly between tools. Much of that work centers around COLMAP, which remains a foundational format across Gaussian Splatting and NeRF workflows.

RealityScan 2.1.1 improves both import and export compatibility with COLMAP by adding support for the FULL_OPENCV camera model and introducing more flexible export structures. Users can now choose between a flat folder layout or the standard COLMAP directory structure, making it easier to plug RealityScan outputs directly into downstream training pipelines like Lichtfeld Studio without restructuring datasets.

Mask handling, a persistent friction point in radiance field preparation, also sees meaningful improvements. The COLMAP exporter now supports AI generated masks and allows users to control naming conventions, including whether to apply the traditional “.mask” suffix. Combined with fixes to ensure masks are actually exported when selected, this update removes several points of failure when preparing masked datasets for training.

On the export side, RealityScan now filters out low accuracy points automatically, helping clean up datasets before they reach optimization stages. For splat based pipelines, this kind of preprocessing can reduce the need for manual pruning.

The update also introduces a new XMP export mode tied to image lists, giving users more control over how camera data is packaged. XMPs can now correspond to either distorted or undistorted images, with selectable coordinate systems, making it easier to align RealityScan outputs with the expectations of different radiance-field training setups.

Beyond exports, several smaller changes improve consistency across automated workflows. CLI commands introduced in 2.1 have been renamed for clarity, and a number of issues affecting COLMAP ingestion, trajectory imports, and intrinsic parameter handling have been resolved. In practice, these fixes reduce edge cases where datasets would silently misalign or fail during batch processing.

As with previous updates, a large portion of the release is dedicated to stability. Mask exports that previously rendered as black images have been corrected, raster artifacts in depth and mask outputs have been eliminated, and several issues affecting LiDAR classification, PLY imports, and camera exports have been addressed.

RealityScan 2.1.1 is available now through the Epic Games Launcher.

Featured

Recents

Platforms

Rapid Splat Renderer 0.9.0 Nearly Doubles Frame Rate with New Rendering Pipeline

The rendering rate of rapid splat rendering increased rapidly.

Michael Rubloff

May 29, 2026

Platforms

Rapid Splat Renderer 0.9.0 Nearly Doubles Frame Rate with New Rendering Pipeline

The rendering rate of rapid splat rendering increased rapidly.

Michael Rubloff

Platforms

Corona 15 Makes Gaussian Splats Participate in Global Illumination

Chaos just announced Corona 15, with gaussian splatting now receiving global illumination.

Michael Rubloff

May 27, 2026

Platforms

Corona 15 Makes Gaussian Splats Participate in Global Illumination

Chaos just announced Corona 15, with gaussian splatting now receiving global illumination.

Michael Rubloff

Platforms

Niantic Spatial and Spexi Announce Partnership for City Scale Gaussian Splats

The two companies announced a partnership for collecting high quality datasets for Physical AI.

Michael Rubloff

May 27, 2026

Platforms

Niantic Spatial and Spexi Announce Partnership for City Scale Gaussian Splats

The two companies announced a partnership for collecting high quality datasets for Physical AI.

Michael Rubloff

Platforms

Manycore Tech Open-Sources Aholo Viewer for Large Scale 3DGS Scenes

Manycore Tech has open-sourced Aholo Viewer, a browser based Gaussian Splatting renderer that uses chunk-level LOD streaming to handle scenes with over one billion splats.

Michael Rubloff

May 26, 2026

Platforms

Manycore Tech Open-Sources Aholo Viewer for Large Scale 3DGS Scenes

Manycore Tech has open-sourced Aholo Viewer, a browser based Gaussian Splatting renderer that uses chunk-level LOD streaming to handle scenes with over one billion splats.

Michael Rubloff

Michael Rubloff

Written by Michael Rubloff

Michael is the Founder and Managing Editor of Radiancefields.com

Email
Copy Link
Twitter
Linkedin
Reddit
Whatsapp

More from Michael Rubloff

More from Michael Rubloff