Support

User Guide

The complete user guide covers importing, training, the viewport, editing, exporting, and keyboard shortcuts.

Frequently Asked Questions

What hardware do I need?

An Apple Silicon Mac (M1, M2, M3, or M4) with macOS 15 Sequoia or later. Intel Macs are not supported because Gaussian Splatting training requires the Metal GPU compute capabilities of Apple Silicon.

How many photos should I use?

For best results, use 10 to 50 photos taken around the object from different angles. Ensure at least 60% overlap between consecutive images. More images generally produce better quality but increase processing time.

Can I use video instead of photos?

Yes. Drop an MP4 or MOV file into the app and RadianceKit will automatically extract frames. You can adjust the sampling density to control how many frames per second are used.

How long does training take?

It depends on the number of images, resolution, and the quality preset. A Balanced preset (20,000 iterations) on 30 images typically takes 2 to 5 minutes on an M1 Pro. The Quality preset takes roughly twice as long.

What is the difference between Classic and MCMC?

Classic densification uses the original clone/split/prune approach from the 3D Gaussian Splatting paper. It is faster and well-tuned. MCMC (Markov Chain Monte Carlo) uses Stochastic Gradient Langevin Dynamics from the NeurIPS 2024 paper. It produces roughly 70% fewer Gaussians, resulting in smaller file sizes, but takes longer to train.

My scene has floating artifacts. How do I fix that?

Use the built-in Gaussian editor. Press Tab to enter edit mode, paint-select the artifacts with the brush, and press Delete to remove them. You can undo with Cmd+Z.

Report a Bug

Please include:

A description of the problem
Steps to reproduce it
Your macOS version and Mac model
The number of images or video length you used

Request a Feature

Describe what you would like RadianceKit to do and how you would use it.

Contact

For other questions, reach out at info@kindler-dev.de.