Blender previs + Seedance 2.5: a new workflow for directing AI video

CurieuxExplorer · x · 2026-09-22

A creator demonstrates a new AI video workflow: build the full camera move and scene structure in Blender (a Statue of Liberty orbit), then feed the render into Dreamina's Seedance 2.5 as reference. Seedance 2.5 impressively understood the Blender model — statue, island layout, orbit, perspective, scale, spatial structure — turning rough 3D previs into cinematic, realistic footage. Pipeline: Astra → Blender → Seedance 2.5, with Blender handling geometry/motion and Seedance handling realism — arguably a new way to direct AI video.

Related event: Blender Previs + Seedance Workflow Goes Viral for Cinematic AI Video(2 posts)→

Original post →

More from Multimodal

Multimodal channel →