Locking geometry and camera first: a GPT + Blender + Seedance AI video pipeline

alifcoder · x · 2026-09-08

@HeyNayeem shares an AI video workflow: GPT-6 Astra generates scene geometry, Blender locks camera and layout, Dreamina's Clay Renderer plugin exports the clay model, and Seedance 2.5 renders final footage. The key insight: Seedance doesn't guess the scene from scratch — geometry, camera and composition are fixed in advance, so it focuses purely on generation. A commenter adds that locking geometry and camera first makes the whole pipeline far more predictable.

Related event: GPT-6 Astra + Blender + Seedance 2.5 Pipeline Turns Code into 3D Video(13 posts)→

Original post →

More from Multimodal

Multimodal channel →