DuplexCascade: Full-Duplex Speech-to-Speech Dialogue with VAD-Free Cascaded ASR-LLM-TTS Pipeline and Micro-Turn Optimization

DuplexCascade introduces a VAD-free cascaded streaming pipeline that enables full-duplex speech-to-speech dialogue while maintaining LLM intelligence. The system converts traditional long utterance turns into micro-turn interactions using special control tokens to coordinate turn-taking and response timing.

• →DuplexCascade solves the trade-off between full-duplex interaction and conversational intelligence in speech dialogue systems.
• →The system eliminates Voice Activity Detection (VAD) segmentation that typically forces half-duplex communication.
• →Micro-turn optimization breaks down conversations into smaller chunks for rapid bidirectional exchange.
• →Special conversational control tokens help coordinate LLM behavior under streaming constraints.
• →The system achieves state-of-the-art performance on Full-DuplexBench and VoiceBench benchmarks.