RealityScan, one of the leading Photogrammetry platforms, often used for its Structure from Motion (SfM) pipeline, just got a major update. At Epic Games’s UnrealFest in Orlando, it has been announced that RealityCapture will be rebranded to go underneath the RealityScan name. There will also be RealityScan Mobile for a simpler version to create textured models.
However, there are also some incredible features coming to the SfM powerhouse. This release introduces several enhancements with significant implications for the future of radiance field reconstructions and offers a glimpse into how Epic Games is thinking about the long term role of this indispensable tool.
Outside of the branding update, the most notable update is a smarter alignment system. RealityScan now defaults to a higher-quality feature detection mode, delivering tighter camera alignments with fewer disjointed components. For Gaussian Splatting workflows, that matters more than it might seem. Better pose accuracy from the start means fewer ghosting artifacts and less need for aggressive pruning. The alignment process now also benefits from GPU acceleration, speeding up initial solves, though it still runs on CPU when needed, so lower-end setups aren’t left behind.
Another standout is the addition of AI-based masking. Anyone who’s ever spent hours creating background mattes for turntable or object-flip shoots will appreciate being able to stay inside RealityScan and use its built-in segmentation model. It cleanly separates subjects from their backgrounds and lets you iterate masks without the usual export/import dance. The RealityScan team recommends having at least 10GB of VRAM for GPU acceleration.
For those who want more control over quality, a new analysis mode lets you visualize sparse point cloud and mesh fidelity using intuitive color overlays. Green means well-covered; red means you might need to go back and shoot from a different angle. You can even bake this information into your textures, making it easier to identify weak spots before committing to a full render or splat conversion.
And for those working with aerial scans, RealityScan 2.0 now includes native support for aerial LiDAR files. You can import common formats like LAS, LAZ, and E57. That means unordered point clouds from drones can now slot directly into your pipeline. For large-scale radiance field projects, this opens the door to hybrid workflows where color information from gaussian splatting to be paired with the accuracy of the LiDAR scans. This also means that RealityScan will be able to ingest XGRIDS data for exterior captures, allowing for a new reconstruction pipeline to the hardware manufacturer's LCC Studio. Fusion with photogrammtery with aerial or terrestrial lidar.
Since April of last year, RealityScan has been free to use for individuals or companies making underneath a threshold. I strongly recommend people learn to use RealityScan as part of their gaussian splatting or other radiance field workflow. There is no formal release date for the 2.0, but it appears that it will make its way to Epic Games Launcher before the end of the month.
You can learn more about RealityScan 2.0 and download the current RealityCapture 1.5 from the Epic Games Launcher.
