LatentSync

LatentSync is an end-to-end lip-synchronization framework jointly launched by ByteDance and Beijing Jiaotong University. It is based on audio-driven latent diffusion models and aims to achieve seamless temporal consistency and generate high-quality, realistic speaking videos. The framework is suitable for a variety of application scenarios such as voice-overs, virtual avatars, game development, and more. LatentSync Features End-to-end lip sync: Latent Sync eliminates the need for any intermediate motion representation, directly in the latent space...