AI Agents Now Control Blender Directly Via MCP, Lowering 3D Barrier
This integration of LLM agents with Blender via MCP could make 3D modeling accessible to non-experts and accelerate previs work for AI-driven video tools like Seedance 2.0.
Reporting from 1 source: ASCII.jp.
A new method using Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol via Codex to control Blender through the MCP protocol is gaining attention. The technique lets users give text instructions in Japanese to generate 3D models and previs videos, dramatically lowering the learning curve for 3D creation. The article demonstrates with examples including a summer Japanese room scene used as reference for AI video generation.
An ASCII.jp feature details how Claude Fable 5 and GPT-5.6 Sol can control Blender through the MCP protocol, reducing the need for manual 3D modeling skills. The method uses a new MCP Connector released by Blender developers in April, which allows the LLM client to handle environment setup and model generation from simple text prompts. The article walks through creating a summer Japanese room scene as a previs video, then feeding it into Hailuo AI's Seedance 2.0 for final video output. AI engineer posi_posi has proposed a workflow that starts with image generation in GPT Image 2.0, then refines the model step by step through Blender MCP commands.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.
Sources
- ASCII.jp AIがBlenderを勝手に操作 3D制作のハードルが一気に下がった