Voice Agent Preambles
Voice agent preambles are short spoken status updates that let a realtime agent acknowledge a user request, explain what it is doing, and keep the user informed while reasoning or tool calls take time.
Key points
- OpenAI's GPT Realtime 2 demo says preambles become more important when voice agents have reasoning and parallel tool calling, because tool actions can take a few seconds [src-051].
- A preamble lets the model communicate during reasoning and tool execution, avoiding dead air while still keeping the user in the loop [src-051].
- In the demo, the assistant says it will pull context and update the CRM before returning customer context and blockers, which is the practical preamble pattern [src-051].
- Preambles fit voice UX especially well because the user cannot see background logs; the spoken update is the interface's equivalent of a progress indicator [src-051].
- The Build Hour shows preambles as part of GPT Realtime 2's reasoning behavior: the model can say it is checking something, keep working across tool calls, and then return a concise result instead of leaving silence [src-083].
- In production, preambles sit beside turn-taking controls. Sierra notes that some moments, such as mandatory disclaimers, may need non-interruptible VAD behavior, while ordinary reasoning should remain interruptible and conversational [src-083].
Related entities
Related concepts
Source references
- [src-051] OpenAI – "We’re introducing three audio models in the API" (2026-05-07)
- [src-083] OpenAI – "Build Hour: GPT-Realtime-2" (2026-05-13)
Recommended next
Keep reading from this thread
From 491 indexed pages and articles.
- Wiki concept GPT Realtime 2 An OpenAI realtime audio model for voice agents that can follow instructions, reason, preserve conversational context, call tools, operate Related by 051
- Wiki concept Voice Agents AI agents whose primary interface is spoken conversation over a phone call, app, or in-browser audio stream. Related by 051
- Insight Recommendation Systems in Production How recommendation systems become production decisioning systems through signals, ranking, constraints, feedback loops, and experimentation Readers have engaged with this next