Structured AI video pipeline: GPT-6 Astra for geometry, Blender scenes, Seedance 2.5 render
Div_pradeep · x · 2026-09-09
A practitioner shares a structured alternative to prompt-regenerate AI video: GPT-6 Astra generates coded geometry, Blender sets up the scene, a Clay Renderer plugin bridges to Dreamina, and Seedance 2.5 produces the final render. The key idea is building a deterministic scene skeleton first, giving far better shot control when composition matters.
More from Multimodal
- Matt Turck generates an Anthropic S-1 explainer video with a single unedited prompt — mattturck · 2026-10-27
- Local text-to-video trailer: Maestro on a 3080Ti plus Premiere Pro Beta's AI Assistant auto-editing — cocktailpeanut · 2026-10-27
- Gen games don't need to be realtime — turn-based interactivity works, devs argue — djcows · 2026-10-27
- Lip sync face-off: Omniflash vs Minimax H3 on The Everlight — Uszko_Visuals · 2026-10-27
- Short film 'Odissey of Aeons' made with MiniMax H3 video generation — Musikalandia · 2026-10-27
- Remiqora Video: open-source local studio turns one idea into 4K chained AI videos — inikolax · 2026-10-27