X-AVDT: Audio-Visual Cross-Attention for Robust Deepfake Detection | ScienceToStartup | ScienceToStartup