MCP for Blender
Menu generate_3d

Tools

generate_3d: Generate 3D Models from Text or Images

generate_3d creates a textured 3D model from a text prompt or an image with Meshy, Tripo, Hunyuan3D or Hyper3D Rodin, and imports it into Blender in one call.

generate_3d creates one textured 3D object from a text prompt or an image and imports it into your scene. It works with Meshy, Tripo, Hunyuan3D and Hyper3D Rodin, and it handles submitting, waiting and importing in a single call.

Choosing a generator

With provider="auto" (the default), the generator is picked in this order:

  1. Generators switched on in MCP for Blender Premium: Meshy, then Tripo, then Hunyuan3D, then Hyper3D Rodin. No API keys needed.
  2. Generators you've set up with your own API keys: Hunyuan3D, then Hyper3D Rodin, then Meshy.

Pass provider to use a specific one.

Parameters

Parameter Description
prompt A short English description of one object, like "weathered wooden treasure chest".
image Instead of a prompt: an absolute image file path or an http(s) URL.
name Object name in the scene. Defaults to one made from the prompt.
provider auto (default), meshy, tripo, hunyuan3d or hyper3d.
quality standard or high (Premium). Leave it out to use your default from the Premium preferences. Not used by Meshy.
bbox_condition Hyper3D Rodin only: [length, width, height] proportions.
job A handle from an earlier call, to keep waiting for it.
wait_seconds How long to wait before returning (10–600, default 50).

Give either prompt or image, not both.

Long generations

Generation usually takes 1–3 minutes. If it isn't done within wait_seconds, generate_3d returns a job handle, and the AI calls it again with job to keep waiting. Once the model is ready, it's imported automatically.

Good practice

  • Generate hero and custom objects, one at a time. Don't generate a whole scene, the ground, or parts to assemble.
  • Generate an object once and duplicate it for repeats. Each call uses a monthly generation or costs money on your own API key.
  • After import, use the reported world bounding box to fix the scale and put the object on the ground (Meshy models already arrive at real-world size, standing on the ground), then check it with look.

Images attached in the chat can't be passed to Blender. Give the AI a file path or a URL instead.