FLOAT: Generative Motion Latent Flow Matching for Audio-driven Talking Portrait

Open in new window