Summary
Zinho Automates demonstrates how GPT-6 Astra, running inside ChatGPT Codex, can take autonomous control of Blender — the open-source 3D animation software — to build, animate, and export complete scenes from text prompts alone. The three-tool chain works as follows: Astra handles scene logic and camera choreography inside Blender, blocking out shapes and motion paths, then Higgsfield applies photorealistic rendering over the rough geometric skeleton.
Five use cases are covered in sequence: a dragon flying through a canyon with a cinematic chase camera, a product advertisement requiring physics-accurate timing (a cookie snapping on beat with chips frozen mid-air), a seamless multi-world continuous camera journey of the kind seen in music videos and trending Reels, a full game trailer built from concept art generated in Higgsfield, and a product visualization sequence. The key insight throughout is that Astra does not attempt detailed modeling — it uses simple primitives like cylinders and spheres — because Blender’s role is motion and camera precision, not aesthetics. That constraint makes the workflow reliable; simple shapes rarely break.
Setup involves signing up for Higgsfield, connecting it in Codex via the plugins menu, and installing the Higgsfield Blender plugin. Once authorized, Astra operates Blender autonomously for the remainder of the session. All prompts are free in the creator’s community school.
📺 Source: Zinho Automates · Published September 12, 2026
🏷️ Format: Hands On Build
