印刻万物 TOP3DGS印刻万物TOP3DGS

Extended notes · Tooling

Getting started with Gaussian Splatting: five mainstream tools

A rewritten Radiance Fields guide comparing Scaniverse, Polycam, Luma AI, KIRI Engine, and Postshot across onboarding, cloud dependence, export capability, and local training trade-offs.

Cross-checked against public sources

Start by sorting by barrier to entry

The guide separates Gaussian Splatting onboarding into mobile capture, web cloud generation, higher-quality cloud processing, professional mesh conversion, and desktop local training. Scaniverse and Polycam Web are good first experiments; Luma AI and KIRI Engine serve users willing to wait for cloud processing and richer export paths; Postshot fits users with NVIDIA RTX hardware who want local control.

Cloud convenience still depends on export

Polycam, Luma AI, and KIRI Engine reduce setup and installation work, but production use depends on export formats, subscription limits, and post-processing support. Cloud workflows are faster for web display or sharing; archival, engine delivery, or retraining workflows should keep original imagery and master files such as PLY.

Where local training fits

Postshot keeps training on the desktop, which helps with data control, parameter iteration, and log inspection. The trade-off is hardware requirement, VRAM budget, and debugging time. Beginners should first learn how capture quality affects results through lower-barrier tools, then decide whether a local GPU workflow is worth the investment.

Related learning path

understand-gaussian-splatting · Module 05

Sources