
by OpenAI
OpenAI's Realtime API lets developers build speech-to-speech voice agents that work directly with audio, keep conversation state and call tools. It is for developers building voice assistants and other low-latency conversational apps.
APIUsage-based
Step 1: Create an OpenAI API key in the API dashboard.
Step 2: On your server, create an ephemeral client secret with POST /v1/realtime/client_secrets.
Step 3: In your frontend, define a RealtimeAgent with instructions using the Agents SDK (@openai/agents/realtime).
Step 4: Open a RealtimeSession with a model such as gpt-realtime-2.1 and connect using the ephemeral key.
Step 5: Attach tools, handoffs and guardrails to the agent, and tune voice activity detection for turn-taking.
It is billed per token, with separate rates for audio and text and for input and output. OpenAI's pricing page lists the current rates for each model.
Yes. Browser apps typically connect over WebRTC, and server-side apps can connect over WebSocket.
Yes. Realtime sessions can call functions, MCP servers and connectors while the conversation continues.
Remove the OpenAI-Beta: realtime=v1 header, create ephemeral credentials with POST /v1/realtime/client_secrets, use /v1/realtime/calls for WebRTC, and update session and event shapes to the GA format.
Environment variables
ChatGPT SpaceChatGPT Space is a shared place in ChatGPT for pages, files, spreadsheets and slides that you build with ChatGPT and your teammates. It replaces Library and is available on Pro, Business and Enterprise plans.
ChatGPT WorkChatGPT Work is an agent in ChatGPT that takes action across your apps and files and turns a goal into finished work such as sheets, slides, docs and web apps. It can stay with a project for hours.
OpenAI DotsDots are always-on agents in ChatGPT that take on ongoing responsibilities and keep making progress between conversations. You choose which apps they use and when they need your approval.
AssemblyAIAssemblyAI is a voice AI platform with APIs for speech-to-text, text-to-speech, speech understanding and LLM access. Developers use it to build voice agents, notetakers and call analytics.
Resemble AIResemble AI detects AI-generated audio, video and images and offers watermarking and identity tools. It is used for fraud prevention, meeting security and content moderation, through an API or built-in integrations.
GladiaGladia is an audio infrastructure API that transcribes speech in real time or from uploaded files and adds audio intelligence such as sentiment and entity detection. It is aimed at developers building voice products.