Learning to Generate Conditional Tri-plane for 3D-aware Expression Controllable Portrait Animation

Researchers present Export3D, a new AI method for creating 3D-aware portrait animations from a single image with controllable facial expressions and camera angles. The technique uses a tri-plane generator and contrastive pre-training to avoid unwanted appearance changes when transferring expressions between different identities.

• →Export3D enables one-shot 3D portrait animation with expression and camera control from single images
• →The method uses tri-plane generation instead of traditional image warping for better appearance-expression separation
• →Contrastive pre-training framework prevents unwanted appearance swapping during cross-identity expression transfer
• →The approach generates 3D-aware images through differentiable volume rendering
• →Research advances controllable portrait animation without relying on heavy image warping techniques