OpenAI Realtime Console is an open-source React application published by OpenAI for inspecting, building and debugging applications on top of the Realtime API, the interface behind low-latency, speech-to-speech interactions with models such as GPT-4o. Rather than wiring together your own test harness, you can open the console, connect it to the API and watch a live voice or text session unfold event by event. The project has collected about 3,600 stars on GitHub and was still receiving updates in 2025.
The console connects to the Realtime API through a small local relay server (or ephemeral tokens) and visualizes everything that happens in a session: conversation items, audio buffers, function calls, interruptions and server events are all logged in real time. Developers can type or speak to the model, inspect the exact JSON exchanged, tune voice activity detection and turn detection settings, and experiment with instructions and tools before committing to a production design. Because it is a complete working client rather than a mock-up, its source code doubles as a reference implementation showing how to manage WebRTC/WebSocket audio, stream transcripts and handle tool calls in a browser-based voice agent.
The console itself is completely free and open source under the MIT license. You only pay for your own OpenAI API usage while testing.
OpenAI Realtime Console is for developers building voice agents, realtime assistants or speech-driven interfaces on the OpenAI Realtime API who want a transparent debugging and prototyping workbench.