Skip to main content
Runchat ships with several image-to-3D model generators. This example covers when to use which, how to produce higher-quality output with multi-view inputs, and how to import the result into your CAD tool. Time: 6 minutes You’ll need: An image to convert (an AI-generated render or a photograph)

Open the example workflow

Make a copy of this workflow in your account to follow along.

​
Pick the right model

Open the Add Nodes menu and scroll to Create 3D Model. The main options:
ModelCostBest for
Trellis~50 creditsSimple, known objects (characters, plants, basic forms). Cheap and fast.
Trellis 2~750 creditsBetter quality, supports multiple input images automatically.
Hunyuan 3D~1,000 creditsHighest quality. Supports explicit multi-view inputs (front, back, left, right).
Marble Mini~500 creditsGaussian splats from a single image. Useful for environments rather than single objects.
For known object types (a chair, a tree, a lamp), Trellis does a surprisingly good job for the price. For anything experimental or architectural, the cheap models tend to fall apart, switch to Hunyuan or Trellis 2. See Credits for the full pricing table.

​
Single image to 3D

The basic workflow:
  1. Drag from an image onto the canvas
  2. Add a 3D model generator (start with Trellis to test)
  3. Run
Trellis outputs a GLB. You can import it directly into Rhino, Blender, Revit, or SketchUp using the Runchat plugins.

​
Multi-view for better results

Single-image 3D fails on complex or experimental geometry, the model has to invent everything it can’t see. The fix is to generate additional views first and feed all of them in.
  1. Take your generated image.
  2. Use Flux 2 Turbo to generate three more views with simple prompts:
    • “Show this building from behind”
    • “From the right side elevation”
    • “From the left side elevation”
  3. Add a Hunyuan 3D node and publish the back, left, and right image parameters so they appear as inputs.
  4. Connect the four views to their corresponding inputs.
  5. Run. Takes a few minutes, around 1,000 credits.
Trellis 2 also accepts multiple images and figures out the multi-view mapping itself, which is less work to wire up. Try it before going to Hunyuan if your geometry isn’t too complex.

​
Why generate 3D from a 2D concept

You’re not producing construction-ready models. You’re taking something visually interesting from a 2D generation, a form, a proportion, a detail like a scalloped staircase or an unusual opening, and getting it into 3D so you can model against it. Reference geometry for a design process, not a deliverable.

​
Import into your CAD tool

For Rhino: select the 3D node and click Import in the action bar of the Runchat for Rhino plugin. The textured model loads at the origin as real Rhino geometry. For SketchUp, Blender, or Revit: same pattern via their respective plugins.

​
Next steps