NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities

AI Digest

NemotronLabs VoiceChat是首个开源全双工语音模型，集成工具调用能力，通过流式处理实现对话交互，测试表现优异但仍有优化空间。

NemotronLabs VoiceChat introduces an open full-duplex speech model with tool calling, achieving strong benchmark results while facing accuracy challenges.

Key points

• 融合流式语音编解码与并行输出流，支持实时对话交互 Combines streaming speech codecs with parallel output streams for real-time interaction
• 具备结构化函数调用与增量转录功能，增强工具集成能力 Supports structured function calls and incremental transcription
• 在Full-Duplex-Bench测试中表现优于多数开源系统 Outperforms most open systems in Full-Duplex-Bench tests
• 需提升论点准确性与端到端工具执行效率 Needs improvement in argument accuracy and tool execution

Takeaway: 语音交互与工具调用的深度融合推动了全双工对话系统的发展。 / Integration of speech interaction and tool calling advances full-duplex dialogue systems.

Why it matters 展示语音交互与工具结合的创新路径，对实际应用有启发价值。