OpenAI: GPT-Live 1
OpenAI full-duplex voice model for natural, expressive speech-to-speech conversations that listens while speaking and supports client delegation to your application.
Providers
Different providers can host the same model. Choose one provider when you need a fixed backend, or let routing select among them.
Supported languages
No reviewed language list for this model yet.
Features
Full-duplex speech-to-speech over WebSocket, listening while it speaks.
speech_to_speech
Accepts text input and returns text output alongside audio.
text_input_output
The model manages when to listen and speak over a continuous audio stream, with interruption handling.
turn_taking
Streams timestamped transcripts of both input and output speech.
live_transcripts
Emits client delegation events for your application to handle and accepts application-provided results back into the conversation.
client_delegation
Code
Live speech-to-speech with the OpenAI SDK
Opens a GPT-Live session, streams question.pcm (24 kHz 16-bit mono PCM) in real time and saves the spoken reply to reply.pcm.
import { readFile, writeFile } from "node:fs/promises";
import { setTimeout as sleep } from "node:timers/promises";
import OpenAI from "openai";
import { LiveWS } from "openai/resources/live/ws";
const client = new OpenAI({
apiKey: process.env.ALLMODELS_API_KEY,
baseURL: "https://api.allmodels.io/oai"
});
const live = new LiveWS(client);
const reply = [];
let quiet;
live.on("session.output_audio.delta", ({ delta }) => reply.push(Buffer.from(delta, "base64")));
live.on("session.output_transcript.delta", () => {
// End the session two seconds after the reply's transcript goes quiet.
clearTimeout(quiet);
quiet = setTimeout(() => live.send({ type: "session.close" }), 2000);
});
live.on("session.closed", async () => {
await writeFile("reply.pcm", Buffer.concat(reply));
live.close();
});
live.on("error", console.error);
live.send({
type: "session.start",
session: {
model: "gpt-live-1",
audio: { format: { type: "audio/pcm", rate: 24000 }, output: { voice: "marin" } },
delegation: { type: "client" }
}
});
// Stream 100 ms frames in real time, then two seconds of silence.
const audio = Buffer.concat([await readFile("question.pcm"), Buffer.alloc(2 * 24000 * 2)]);
for (let offset = 0; offset < audio.length; offset += 4800) {
live.send({ type: "session.input_audio.append", audio: audio.subarray(offset, offset + 4800).toString("base64") });
await sleep(100);
}