[::] Agentic Stacks
← Tools · Voice Infrastructure
{ } JSON

OpenAI Realtime API

used in
0
stacks

Speech-to-speech realtime model API for building low-latency voice agents.

commercial usage https://developers.openai.com/api/docs/guides/realtime ↗ 
Exposes the gpt-realtime model for speech-to-speech agents that listen, reason, speak, and call tools over persistent low-latency sessions. Supports WebRTC for browser and mobile clients, WebSocket for server-side media pipelines, and SIP for telephony. Specialized variants handle continuous live translation and streaming transcription. Billed usage-based on audio and tokens through the OpenAI platform.

Strengths

  • Single model handling speech-to-speech with reasoning and tool calling
  • Multiple transports including WebRTC, WebSocket, and SIP for telephony
  • Low-latency persistent sessions for natural turn-taking
  • Specialized variants for live translation and streaming transcription

Tradeoffs

  • Locked to OpenAI models with no provider choice
  • Closed, cloud-only API
  • Realtime audio pricing can be costly at scale

Get the weekly digest

New stacks, new tools, what's moving. No spam.