Orchestration · Software component

Voice Turn Coordinator

Software componentOrchestrationOrchestration & Toolsarc:VoiceTurnCoordinator

A runtime component that sequences one spoken conversation turn: streams user audio to speech recognition, hands transcripts to the agent, and passes the agent's reply to speech synthesis.

Responsibility. Coordinates the ASR-agent-TTS turn pipeline within the conversational latency budget.

Also known as: Voice agent pipeline, Voice interaction orchestrator

Direct neighbourhood (hover for relationship types)

Relationships

invokes dependency

Agent Controller abstract Ch7.5
Automatic Speech Recognizer abstract Ch7.5
Speech Synthesizer Ch7.5

receives data from dynamic

Design guidance

SHOULD target end-to-end voice turn latency near human turn-taking (~200-400ms) and avoid >1 second.
SHOULD start agent context retrieval on partial transcripts while the user is still speaking.

Quantitative guidance

As stated by the sources; verify before use.

Users expect <300ms end-to-end; a simple query totals ~300-900ms across ASR, LLM (100-500ms) and TTS (100-200ms) (Ch7.5).

Classification

Patterns: User speaks -> ASR -> LLM agent -> TTS -> user hearsProcess-before-utterance-ends with provisional transcripts
Technologies: NVIDIA Riva
Quality attributes: Performance efficiency (ISO/IEC 25010)
Risks mitigated: End-to-end latency >1s making voice agents feel sluggish

Sources

Ch7.5: T. Nguyen, "NeMo Curator, Riva Speech AI & Multimodal Integration," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 7.5. ISBN: 9798244538229.