Openai Realtime

Low-latency speech-to-speech over a WebSocket, drop-in compatible with the OpenAI Realtime GA API. Connect with wss://api.acedata.cloud/v1/realtime?model=gpt-realtime and an Authorization: Bearer <token> header (browsers pass the token via the Sec-WebSocket-Protocol subprotocol). Client and server exchange OpenAI Realtime JSON events (session.update, input_audio_buffer.append, response.create, response.output_audio.delta, response.done, …). Audio is pcm16 @ 24kHz mono. Billed per assistant turn from response.done.usage (audio and text tokens priced separately).