RadianceKit

What is RadianceKit?

RadianceKit is a native Mac app for 3D Gaussian Splatting. It lets users import photos or video, reconstruct a scene locally on Apple Silicon using Metal GPU compute, then preview, edit, and export the result in one workflow.

The app is positioned as a complete alternative to toolchains that rely on cloud processing, Python, or separate alignment and export steps. It is designed for people who want to create photorealistic 3D scenes on a Mac while keeping the data on their own machine.

Key Features

• Native Apple Silicon GPU training — RadianceKit runs Gaussian Splatting training on M1, M2, M3, and M4 Macs using Metal, so processing happens locally instead of in the cloud.
• Single-app workflow — Import media, align camera positions, train the model, inspect the reconstruction, and export from the same app without a command-line setup.
• Simple and Expert modes — Simple Mode provides a guided flow for quick starts, while Expert Mode adds a project navigator, interactive 3D viewport, live loss curves, and detailed training and export controls.
• Interactive Gaussian editor — Users can select and delete regions directly in the 3D view with a brush tool, which helps remove unwanted artifacts or parts of a reconstruction.
• Multiple export options — Scenes can be exported as PLY, Compressed PLY, SPZ, glTF, .splat, or SOG, with options to create orbit videos or self-contained web viewers.
• Local processing and privacy — The app processes scenes on the user’s Mac and does not require a cloud upload for training.

How to Use RadianceKit

A typical workflow starts by opening the app and importing either photos or a video. RadianceKit then aligns the cameras, trains the Gaussian Splatting scene on the Mac’s Apple Silicon GPU, and lets the user inspect the reconstruction in real time.

From there, users can switch between a guided mode for straightforward jobs or an expert layout for parameter tuning and detailed inspection. When the scene is ready, they can clean up the model with the editor and export it in a format suited to sharing, archiving, or web viewing.

Use Cases

• Photogrammetry-style scene reconstruction on Mac — Convert ordinary photos or video into an explorable 3D scene without leaving the Mac.
• Local 3D capture workflows — Run reconstruction on-device when cloud upload is not desired or when keeping source media local matters.
• Cleanup of reconstructions — Remove floating artifacts, stray objects, or other unwanted regions using the in-app brush editor.
• Technical experimentation and parameter tuning — Use Expert Mode to inspect training curves and adjust settings for more controlled reconstruction work.
• Sharing interactive outputs — Export a scene as a web viewer or orbit video for presentation and review without requiring extra software from the viewer.

FAQ

• Does RadianceKit require cloud processing? No. Training runs locally on the Mac using Apple Silicon GPU compute through Metal.
• Do I need Python or command-line tools? No. The app is presented as a complete workflow with import, training, editing, and export inside one Mac app.
• Which Macs are supported? The source says RadianceKit requires an Apple Silicon Mac (M1 or later) running macOS 26 Tahoe or later, and recommends 16 GB of RAM for larger scenes.
• What kinds of files can it import? It can import videos such as .mp4, .mov, .m4v, and .avi, as well as common photo formats including .jpg/.jpeg, .png, .heic, .tiff/.tif, and .bmp. It also opens existing splats like .ply, .spz, and .splat, plus scene and camera bundles.
• Does it support 360 footage directly? Not yet. The page says Insta360 .insv and stitched 360 equirectangular footage are not directly supported at this time.

Alternatives

• Cloud-based NeRF or photogrammetry services — These can handle reconstruction without relying on local GPU compute, but they require uploading media and using an online processing workflow.
• Desktop photogrammetry applications — Traditional photogrammetry tools focus on image-based reconstruction and may follow different alignment and export workflows than Gaussian Splatting.
• Python-based Gaussian Splatting toolchains — These may offer more low-level control for technical users, but they require setup outside of a packaged Mac app.
• Other 3D capture and reconstruction apps — Adjacent apps may prioritize capture, scanning, or editing rather than a complete Gaussian Splatting pipeline from import through export.