@TensorRT

AVTR-1, an open-weight flow-matching transformer for audio-driven avatars, has been released for real-time dialogue generation. The model renders lip-synced speech and active listening at 25 frames pe…