Skip to content

Preparing models

Upload GLB (recommended) or glTF. Export presets from Blender, 3ds Max, Cinema 4D and SketchUp all work; keep materials PBR (the glTF standard).

Every upload runs through the optimization pipeline. You never configure this:

  • Geometry compression for fast web delivery.
  • Texture encoding — color maps and detail maps are re-encoded to GPU-friendly formats sized for their role.
  • AR derivatives — iOS (USDZ) and Android formats are generated per model, so AR works the moment the project is published.
  • Versioning — replacing a model updates every published surface (viewer, embeds, AR) consistently.
  • Real-world scale. Model in meters at true size — AR places the product at exactly the size you export.
  • Reasonable budgets. Web viewers are not render farms; keep meshes sensible (typically well under 1M triangles) and textures ≤ 2048px where possible — the pipeline compresses further, but starting lean keeps quality headroom.
  • Clean origin. Place the model at the scene origin, resting on the ground plane.

If the product will be configured from separate parts (body + legs + cushion), export each part as its own model in a shared coordinate frame — every part positioned where it belongs relative to the assembled product. The configurator assembles parts by their exported frames, so consistency here is what makes assemblies line up perfectly. Details: Configurator.