AIHacker Noonabout 8 hours ago
NVIDIA VoiceChat-11B Enables Real-Time Full-Duplex AI Speech
12 min read
NVIDIA released VoiceChat-11B, an 11-billion parameter full-duplex speech model with a unified architecture. It processes 16 kHz audio input and outputs 22.05 kHz, achieving ~450 ms turn-taking latency with real-time interruption handling and tool calling, bypassing traditional cascaded ASR, LLM, and TTS pipelines.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyNvidia
