How to Use AI (ChatGPT) to Create 3D Models in Blender

AI tools like ChatGPT can generate Blender Python scripts to automate modeling tasks. Instead of manually building every primitive or modifier setup, you can describe what you want in natural language, have ChatGPT generate Python code, and then run it directly inside Blender.

This workflow is useful for:

  • Procedural asset creation (chairs, buildings, landscapes).

  • Automating repetitive modeling tasks.

  • Exploring new ideas quickly through code prompts.

Step 1 — Getting Setup with the Right Tools (00:52)

  1. Install Blender:
    Download and install the latest version from blender.org.
    Ensure Python is bundled with Blender (default).

  2. Create an OpenAI Account:
    Register at chat.openai.com for access to ChatGPT.
    For more advanced scripting, GPT-4 or newer is recommended.

Step 2 — How to Get Up and Running in Blender (01:08)

  1. Open Blender and select Scripting Workspace from the top menu.

  2. This gives you access to:

    • Text Editor (for writing or pasting code).

    • Python Console (to test snippets interactively).

    • Info panel (shows commands Blender executes as you model manually).

  3. Note: Blender uses the bpy (Blender Python API) module to manipulate objects, materials, and scenes.

Step 3 — How to Get Started with ChatGPT (02:10)

  1. In ChatGPT, provide a clear, detailed prompt about what you want to model.
    Example:

    Write Blender Python code that creates a low-poly table with a square top and four legs.

  2. ChatGPT will generate a Python script using bpy.ops.mesh.primitive_cube_add, scaling, and transformations.

  3. Copy the script and paste it into Blender’s Text Editor.

  4. Press Run Script to execute. The model should appear in the 3D Viewport.

Step 4 — How to Use ChatGPT and Blender Together (03:19)

Iterative Workflow

  1. Describe the asset you want (e.g., “a stylized city block with random building heights”).

  2. Run the generated script in Blender.

  3. Evaluate the result — does it look right?

  4. Refine by telling ChatGPT how to adjust (e.g., “make the buildings taller and add a plane for the ground”).

Practical Examples

  • Generate procedural landscapes using noise modifiers.

  • Create random arrays of objects (trees, street lamps).

  • Automate repetitive edits like beveling edges or applying materials.

Limitations

  • ChatGPT doesn’t run inside Blender natively; you must copy-paste scripts.

  • Complex geometry may need manual adjustments.

  • Efficiency: not always optimized code.

Step 5 — GPT-4 and Beyond (07:07)

  • Improved Code Accuracy: GPT-4+ generates fewer syntax errors and understands the Blender API more deeply.

  • Natural Language to 3D Assets: In the near future, expect tighter integrations where you type or say “make a forest scene” and AI generates full geometry, materials, and lighting.

  • Workflow Enhancements: AI will complement—not replace—artists, handling technical setup so creators focus on design.

Example Workflow: Creating a Procedural Fence

  1. Prompt ChatGPT:

    Write Blender Python code to create a fence: a row of 20 vertical cylinders spaced evenly on a plane.

  2. Copy script → Paste in Blender’s Text Editor → Run.

  3. Modify via ChatGPT:

    Make the cylinders thinner, increase the height, and add a cube as a fence base.

  4. Run updated code to refine the result.

3D ModelsFrancesca Tabor