Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment figure
AlphaXiv 中文概览(可滚动查看)