main story Microsoft’s VASA-1 AI Model Can Generate Deepfakes Using a Single Photo and Speech Audio Track 0