Microsoft's VASA-1 Turns Single Image into Hyper-Realistic Deepfake in 170 Milliseconds

You can't download Microsoft Vasa 1, as it has not been made public yet.

Microsoft's latest model, known as VASA-1, is capable of turning a single image into a hyper-realistic deepfake. It can also use audio to achieve lip-sync with lifelike facial behavior and generate natural head movements in real-time.

VASA-1 outputs videos at a resolution of 512x512 and can achieve up to 40 frames per second (FPS). It operates with a latency of only 170 milliseconds on NVIDIA RTX 4090 GPU systems.

Unlike traditional methods that may handle facial features separately, VASA-1 uses a diffusion-based model to generate holistic facial dynamics and head movements. This method considers all facial dynamics, including lip motion, expression, and eye movements, as parts of a single comprehensive model.