SpeakerVid-5M: A Large-Scale High-Quality Dataset for Audio-Visual Dyadic Interactive Human Generation Paper • 2507.09862 • Published Jul 14, 2025 • 49
