# Data preparation workflow

> Complete guide to preparing 3D models and CAD data for real-time applications.

In this document, "dataprep" refers to “data preparation” and to the process of preparing 3D assets from various sources (CAD, DCC) and formats (CAD, mesh, point cloud) for RT3D use. This encompasses several possible stages: importing and converting source assets into appropriate RT3D assets while retaining critical data, geometrically correcting these assets to address conversion imperfections, enriching and enhancing them for dynamic, value-added RT3D experiences, and optimizing them for specific target devices or platforms.

The typical dataprep workflow consists of six phases:

1. **Import**

   Load raw data and convert to tessellated meshes.

   [Guidelines](./import-guidelines): Learn how to import and prepare CAD files for real-time 3D experiences.


   > **Note:**
   >
   > This results in a converted model.

2. **Fix (optional)**

   Fix potential issues in the model until obtaining a geometrically correct model.

   [Guidelines](./import-guidelines#repair): Learn how to repair CAD files.


   > **Note:**
   >
   > This results in a geometrically correct model.

3. **Stage (optional)**

   Enrich/beautify until obtaining a 100% quality, exploitable model.

   [Guidelines](./stage-guidelines): Learn how to enrich models with UVs, materials, and ambient occlusion.


   > **Note:**
   >
   > This results in a master model (LOD0).

4. **Optimize (optional)**

   Optimize or diminish the LOD0 model quality and file size.

   [Guidelines](./optimization-guidelines): Learn how to optimize CAD files for real-time 3D experiences.


   > **Note:**
   >
   > This results in an optimized master model (LOD0).

5. **Generate LODs (optional)**

   Create multiple Levels of Detail for optimal rendering performance at different distances.

   [Guidelines](./lod-guidelines): Learn how to generate LOD chains with aggressive decimation and advanced techniques.


   > **Note:**
   >
   > This results in multiple LOD levels (LOD0, LOD1, LOD2, LOD3+).

6. **Export**

   Push or publish an optimized model!

   [Guidelines](./export-guidelines): Learn how to export optimized models with the lightest file size using GLB with Draco and KTX2 compression.


   > **Note:**
   >
   > An optimized and converted file.