Tutorial — From scan to 3D Gaussian Splat

This step-by-step guide shows how to extract a dataset from your Raven scan, send it to cloud processing, and get back a photorealistic Gaussian Splat. A few clicks — the cloud does the rest.

① Extract the dataset

1. Open your scan (device folder) in Lidar Extractor.
2. In EXPORT, pick the Gaussian Splat tile.
3. Section “① Prepare the dataset” → COLMAP dataset:
   • Perspective images: leave enabled (undistorted images, compatible everywhere).
   • Higher quality (new): 800 images at 2048 px instead of 400 at 1600 — noticeably sharper 3D. Check it for results you plan to publish; extraction and processing take ~2× longer.
4. Click Extract dataset. The …_colmap folder is created in EXPORT_lidar-extrator (posed images + poses + initialization cloud).

② Send to cloud processing

1. Click Cloud process (create a free account on first use — email verification).
2. Pick your dataset in the list (⟳ button to refresh if you just extracted it), then Send.
3. Uploads resume automatically after interruptions; you can run several jobs in parallel and close the app: everything continues server-side.

③ Track and get the result

• Progress shows live (download → camera positions → training → delivery).
• When done, the …_SPLAT.ply file appears in the Clouds tab with a NEW badge.
• Open it with Splat 3D: fly through, crop, clean up, export.

🆕 RayStudio projects (Raven & Raven MAX)

Do you process your scans with RayStudio? Lidar Extractor can now use its Gaussian export directly:

1. In RayStudio, run the Gaussian preparation of your project (a gaussian folder is created inside the project).
2. Open the same project folder in Lidar Extractor.
3. Click Cloud process: a “RayStudio — your project” dataset appears in the list. Send it.

RayStudio's camera positions are reused as-is: no SfM step, roughly 2× faster processing, and the 3D training gets our photorealistic engine. Works with the Raven (2 cameras) as well as the Raven MAX.

Download Lidar Extractor Help & FAQ