VASA is a framework that generates lifelike talking faces of virtual characters with appealing visual affective skills. The premiere model, VASA-1, can produce lip movements synchronized with audio and capture facial nuances and natural head motions. The core innovations include a holistic facial dynamics and head movement generation model and the development of an expressive and disentangled face latent space. The method outperforms previous methods in terms of video quality, facial and head dynamics, and real-time generation. It paves the way for real-time engagements with lifelike avatars that emulate human conversational behaviors. Responsible AI considerations are also discussed.