Skip to content
Beta

These docs cover RhinoArtisan 7.0, which is still in beta. The contents are provisional: features and pages can still be added, changed or removed before the final release.

Organic Modeling & Sculpting

Image to 3D by AI

When the shape you need already exists — in a photograph, a client’s reference, an antique piece, or a capture of your own viewport — Image to 3D reconstructs it as an organic mesh you can edit. It runs in Generative AI’s Studio and offers five models, so you can pick fidelity, control or speed depending on the piece.

How to run it

  1. Open the Generative AI Studio and pick or create a project.
  2. Add the source image: Upload Image for a photo (JPG/PNG), or Capture Snapshot for a clean 1920×1080 capture of the active viewport. Select it in the grid — the selected item is the generation’s input.
  3. Click the mode/model chip and select Image to 3D, then choose a model:
    • Meshy 6 Preview — the most control: topology, target polycount, symmetry, remeshing and texturing in Model Options.
    • Hunyuan 3D 3.1 Pro — the highest geometric detail on ornamental work, with PBR textures.
    • Hunyuan3D v3 — solid quality at a lower cost, good for quick volume tests.
    • Rodin — clean, production-grade meshes that hold up to further editing.
    • Seed3D — fast one-shot reconstruction, faithful to the photo.
  4. Press Send. The job queues, shows as processing in the grid, and resolves on its own — Refresh Project also picks it up.
  5. Inspect the result in the interactive 3D preview, then hover the item and click Import to Rhino: the mesh lands in the active document, the view zooms to it, and the Studio closes so you can start working.

Tips

  • Give the model a clean subject. A single piece, well lit, on a plain background reconstructs far better than a busy scene. If your photo has a distracting background, run BiRefNet v2 in Image to Image first — it cuts the subject out cleanly, even around thin chains and prongs — and use that result as the input.
  • Compare models on the same input. Reconstruction is not deterministic: the same photo can come out best from different models depending on the shape. Since every result stays in the project grid, running two or three models side by side is the quickest way to find the winner.
  • Then make it jewelry. The imported mesh is a starting form: refine it with Sculpt, rebuild key areas as SubD, and scale it to the piece’s real dimensions before building settings around it.