iPhone LiDAR Scanning Comparison: KIRI Engine and Scaniverse

To explore the free capabilities of two leading mobile 3D scanning apps—KIRI Engine and Niantic Scaniverse—I ran a straightforward test using my iPhone. The subject was my woodstove, which offered a variety of scan challenges. The cast iron had intricate textures and patterns, while the soapstone panels were smooth and low in contrast. A matte black stove pipe and red brick chimney added further complexity. This made it a strong candidate for an iPhone LiDAR scanning comparison, with both textured and featureless surfaces. I followed the same scan path with each app, scanning for about 90–120 seconds.

What’s New in KIRI Engine 3.14

KIRI Engine’s latest update, version 3.14, adds meaningful improvements to its Scene Scan feature. It introduces two machine learning models: StableNormal and PromptDA. StableNormal improves the accuracy of surface geometry, even in low light or while moving. PromptDA enhances depth maps by using LiDAR data to guide a depth estimation model.

These upgrades aim to fix common weaknesses in mobile LiDAR scanning—especially on reflective or translucent materials. The update also allows Pro users to export raw datasets for use in 3D software like Nerfstudio and GauStudio. This makes KIRI more appealing to researchers and professionals who want structured, high-fidelity data.

For more details on the KIRI Engine 3.14 upgrade, read this article.

Hands-On Scanning Test

For this test of iPhone LiDAR scanning apps, I used Lidar Scan > Scene Scan in KIRI Engine. The scan was processed locally. The raw data came in at 39 MB, and I exported a 12 MB ZIP file containing an 11 MB .obj and a small .mtl. The free version does not allow exporting 3D Gaussian Splats (3DGS), which are only available through the Pro Plan.

In Scaniverse, I used the Splat mode, which is free and also processed locally. The raw data was 213 MB and the processed model totaled 269 MB. I exported the scan as a 121 MB .ply file and an 11 MB .spz. The .ply format works with most open-source 3D tools, while .spz is a splat-optimized format developed by Niantic. Learn more about SPZ here.

Results and Observations

In this basic test of free tools, a quick visual inspection clearly shows Scaniverse produced a better model. Scaniverse has better accuracy and captured subtle features of the cast iron. Screenshots of both models are included below. KIRI Engine’s results show distortions and a lack of detail.

Despite the shortcoming of KIRI Engine in this simple test of free features, we are excited to explore its premium features next—especially cloud-based processing and Gaussian Splat exports shown in recent reviews. The examples in the linked review are impressive and we look forward to a more comprehensive comparison that includes KIRI Engine’s Pro Plan features. Scaniverse offers its full feature set for free.

KIRI’s Export Options for Advanced Users

We also want to highlight KIRI Engine’s Pro export features. These include the original photos, depth maps, confidence maps, and full camera metadata. The data is structured for use in tools like Nerfstudio and GauStudio. This makes KIRI suitable for not just hobbyists, but also academic and professional workflows.