ComfyUI Separates Actor Motion, Camera, and Style in AI Video

🎨 ComfyUI Separates Motion, Camera, and Style

A production-like workflow has been assembled: Meta SAM 3D Body reconstructs a full 3D body model with pose and facial expressions from regular video, in Blender the director independently sets the camera — angle, lens, trajectory, and ByteDance Seedance 2.5 turns the draft into a final scene up to 30 seconds long with sound and lip-sync.

🌍 The “video -> 3D scene in Blender -> generation in Seedance 2.5” chain brings the classic previz pipeline to AI video: mise-en-scène and timing are checked before paying for final generations.

👤 Locally, ComfyUI and SAM 3D Body are free — the official template is in the Template Library. In Comfy Cloud, GPU work costs ~$0.08 per minute, and Seedance 2.5 is billed separately with a price before generation. A prompt is not the only way to control a scene.

Source 1: https://docs.comfy.org/tutorials/utility/sam3d-body Source 2: https://docs.comfy.org/tutorials/partner-nodes/bytedance/seedance-2-5/workflow