Poster
in
Workshop: Real-Time Multimodal Conversational AI

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

Jagadeesh Balam ⋅ Travis M Bartley ⋅ Edresson Casanova ⋅ Sanjay S Chauhan ⋅ CHEN CHEN ⋅ Zhehuai Chen ⋅ ZIJIA CHEN ⋅ Slyne Deng ⋅ Shalini De Mello ⋅ Mikyas Desta ⋅ Harishchandra Dubey ⋅ Slim Essid ⋅ Nourchene Ferchichi ⋅ Boris Ginsburg ⋅ Mariana G Fuenmayor ⋅ Ke Hu ⋅ Myungjong Kim ⋅ Viacheslav Klimkov ⋅ Seelan L Narasimhan ⋅ Yi-Hui Lee ⋅ Jason Li ⋅ Eileen M Long ⋅ Ameya Mahabaleshwarkar ⋅ Aditya Malte ⋅ Amrita Mazumdar ⋅ Valya Mendelev ⋅ Koki Nagano ⋅ Oluwatobi Olabiyi ⋅ Wookie Park ⋅ Ankita Pasad ⋅ Yifan Peng ⋅ Elena Rastorgueva ⋅ Rajarshi Roy ⋅ Nikhil Srihari ⋅ Yuanhang Su ⋅ Yoshi Suhara ⋅ Viet A Trinh ⋅ Jinhan Wang ⋅ Piotr Żelasko

Project Page [ OpenReview]

Abstract

We introduce NemotronLabs VoiceChat, an open full-duplex speech-to-speech model with native tool-calling capabilities. NemotronLabs VoiceChat combines a streaming speech encoder and decoder-only language model with parallel specialized output streams for agent text and structured function calls, an auxiliary RNN-T branch for incremental user transcription, and a streaming TTS decoder. This design enables the model to listen, transcribe, reason, invoke tools, and speak within a unified streaming architecture while preserving the temporal behavior required for natural conversation. On Full-Duplex-Bench 1.0, NemotronLabs VoiceChat achieves the lowest pause-handling takeover rates among evaluated open-weight systems, 100\% takeover following user interruptions, and a 4.33/5 post-interruption response-quality score. On Full-Duplex-Bench 1.5, it resumes its response after user backchannels in 93\% of cases. NemotronLabs VoiceChat obtains a 55.1 normalized average on VoiceBench and, on Full-Duplex-Bench 3.0 (FDB 3.0), achieves 82.5\% tool-selection F1, while argument accuracy and end-to-end tool execution remain areas for improvement. These results demonstrate that full-duplex interaction, speech recognition and generation, general language capabilities, and external tool use can be integrated in a single open speech-to-speech model without sacrificing real-time conversational behavior.

Chat is not available.