I’ve been fortunate enough to explore radiance fields since early 2022, when I first stumbled upon NVIDIA’s Instant NGP. The moment I realized that all I really needed was my phone and a computer to reconstruct the world around me, my perspective on imaging changed.
While many people were excited about the potential of lifelike 3D, I was also limited by the equipment I had access to. I didn’t own a high end mirrorless camera, so the natural choice for me was to use my phone to capture.
Fortunately, the ecosystem evolved quickly. Over the past four years, I’ve watched capture and reconstruction apps appear one after another, each introducing their own workflows and proprietary pipelines. I’ve spent countless hours capturing with these tools and experimenting with them, likely more than most people. I genuinely enjoy using them and plan to continue doing so.
That said, I’ve always felt there should be a better way to capture high-fidelity data using the device that almost everyone already has in their pocket.
More specifically, I’ve always believed users should be able to capture the world around them once and then have the freedom to choose whichever reconstruction pipeline they want, while retaining full access to their original data.
With that in mind, today I’m releasing the first version of Gaussian SplatKing.
Gaussian SplatKing is a completely free mobile capture app that supports three capture modes: Video, Photo, and LiDAR.
The app gives users direct control over manual camera settings, including shutter speed, ISO, white balance, and exposure compensation. In both Video and Photo modes, SplatKing also captures simultaneously from the 0.5× and 1× lenses, allowing users to collect more coverage and detail in less time.
When using Photo and LiDAR modes, SplatKing automatically generates per-frame quality predictors, making it easier to quickly identify and remove unusable images.
In LiDAR mode, the app also leverages ARKit to visualize camera frustums directly in space, which are color-coded in real time based on this quality predictor. If the system detects weak coverage, users can immediately return to that area and capture additional data before leaving the scene.
Once a capture session is complete, all images, metadata, and additional information are placed directly into a folder within the Files app, allowing users to zip the data and transfer it via AirDrop or any cloud storage platform of their choice.
And because I’m left handed, I also added a left handed interface mode to make accessing controls easier.
This is still very much the first version of the application, but the goal will always remain the same: to help people create the highest fidelity radiance field data possible using the device they already own.
Gaussian SplatKing is available today as a completely free download on the iOS App Store. There are already a few updates coming this weekend that will further simplify data offloading from the phone and continue improving the capture experience.
The app is free to use, but if you would like to support development there is also a donation page. Nothing is expected, and the app will remain free.
I now firmly believe that radiance fields represent a new imaging medium, extending photography and video beyond the limits of traditional 2D capture. If this tool helps even a small number of people begin exploring that future, I’ll consider it worthwhile.
