I Let AI Control My Blender It Made Me Rethink What It Means for Artists

Summary

Grant Abbitt reports on an experiment using Claude AI to operate Blender. He explores whether AI can create 3D art or work more effectively as an assistant for artists. He presents the video as an initial investigation rather than a definitive verdict and notes concerns about AI’s ethical and creative implications.

Tests and results

• Creating a low-poly fox from a reference image: Claude generated a Python script that produced a model resembling the reference. Requests for more depth and detail led to gaps, awkward geometry, merged parts, and inconsistent results. Abbitt found that asking for “more detail” often added polygons without improving the model’s structure.
• Creating a zombie character: Reference images from multiple angles helped, but the model still had anatomical and construction problems, including misplaced thumbs, distorted features, and awkward clothing. Abbitt observed that building characters in separate parts may be more practical than relying on a single broad prompt.
• Generating variations of an existing character: Claude more successfully changed colors, proportions, and some details while largely preserving the original model’s structure. Although the results had flaws, they could be useful for quickly exploring enemy designs for a game.
• Making texture brushes: Claude created stylized brushes and added them to Blender’s asset library. Abbitt considered these among the more useful results, especially because an artist could refine them with further instructions.
• Moving a Rigify character and animations into Unreal Engine: This was the most technically encouraging test. The first export had movement and deformation problems, and troubleshooting took multiple rounds of about 20 minutes each. Claude eventually improved the result and resolved the main issues. Abbitt sees potential for AI to handle technical work such as export compatibility, rigging, retopology, and UV unwrapping, while artists retain creative control over animation and character personality.

Overall assessment

Abbitt concludes that the tools he tested are not ready to replace artists. Direct model creation was inconsistent and often required substantial guidance, while supporting tasks—especially creating brushes and addressing technical workflow problems—seemed more useful.

He suggests AI could help artists spend less time on routine technical work, but warns that automating entry-level tasks could make it harder for beginners to enter the industry and affect jobs.

Reviews, guides, or tutorials

This is a hands-on review and experiment rather than a step-by-step tutorial. The tests covered model generation, character variation, brush creation, and rig export.

Main speaker/source: Grant Abbitt.