Capturing gaussian splatting data with 360 cameras is one of the most enjoyable processes. There’s a certain comfort knowing how much coverage is being captured in a single pass. Where things come quickly to a halt is the post processing step. While it certainly has become easier thanks to things like COLMAP support for 360 cameras, the process can still be foreign to navigate.
That said, Olli Huttunen’s Blender plugin has made it so much easier for me to slice up and render out still images for training and it just received its first update of the year.
Version 1.2.1 introduces support for full 360° image sequences. Instead of relying solely on video input, users can load folders of equirectangular photos, with automatic detection of valid frame sequences and proper frame range configuration based on sequence length. For drone operators and structured capture workflows, this is a meaningful addition. Still image sequences often provide more consistent exposure and metadata integrity than compressed video streams, which directly affects downstream reconstruction quality.
The update also deepens GPS metadata integration. When EXIF data is available, the add on can automatically read latitude, longitude, and altitude values. For outdoor scenes and aerial capture in particular, this strengthens alignment foundations before data ever reaches COLMAP or a splatting trainer. Given the recent release of the Antigravity A1, it’s easy to understand where these updates can become very helpful. While GPS alone does not guarantee perfect reconstruction, initializing scenes with real world coordinates can reduce ambiguity and improve spatial coherence, especially in larger environments.
Media handling has also been refined. Background loading logic for both video and image sequences has been stabilized, frame duration detection has been improved, and world node setup inside Blender has been cleaned up.
The camera group system, one of the add on’s defining features, has received structural improvements as well. The non destructive rotation system is now more stable, camera angle offset logic has been refined, and field of view synchronization across cameras has been improved. If offsets drift or FOV relationships misalign, reconstruction consistency suffers. Tightening that system increases confidence in repeatable dataset generation.
Rendering has been expanded to include Sweeping and Running Order modes, offering more control over how frames are exported. Compatibility fixes for newer Blender versions round out the release, alongside node setup and property handling improvements based on real world testing.
As more creators experiment with using 360 cameras as structured scanners for 3D Gaussian Splatting, making it easier to go from capture to reconstruction is critical.
For those already using 360 Extractor, the update is available through the Gumroad library. For those exploring 360 based splat capture for the first time, you will need to be using Blender 3.6 or higher and it’s available here.
New to Gaussian Splatting? Start here
