Quickstart

The Sync API turns up to 120 seconds of speech into one finished transcript, returned in the same HTTP call. No job to poll, no WebSocket to manage. This page gets you from a microphone to that transcript on the fastest path the API offers. The upload runs while the person is still speaking, so what you wait for at the end is only the last stretch of audio.

Before you begin

An API key. Copy it from API Keys and set it once:
export ASSEMBLYAI_API_KEY=<your-key>
The SDK.
- Python
- JavaScript
pip install -U assemblyai sounddevice
npm install assemblyai
The JavaScript example records with SoX (brew install sox or apt install sox).

Transcribe from a microphone

Open a session, hand it audio as you capture it, and read the result when the speaker stops.

Python
JavaScript

import sounddevice as sd
from assemblyai.sync.v1 import SyncTranscriber, SyncTranscriptionConfig

RATE = 16000
config = SyncTranscriptionConfig(sample_rate=RATE, channels=1)

with SyncTranscriber() as transcriber:
    with transcriber.open_live(config) as session:
        microphone = sd.RawInputStream(
            samplerate=RATE, channels=1, dtype="int16",
            callback=lambda data, *_: session.write(bytes(data)),
        )
        with microphone:
            input("Speak, then press Enter... ")

    print(session.result().text)

import { spawn } from "node:child_process";
import { AssemblyAI } from "assemblyai";

const client = new AssemblyAI({ apiKey: process.env.ASSEMBLYAI_API_KEY });
const RATE = 16000;

const session = client.sync.openLive({ sample_rate: RATE, channels: 1 });

// SoX writes raw 16-bit mono PCM to stdout while it records.
const microphone = spawn("sox", [
  "--default-device", "--no-show-progress",
  "--rate", String(RATE), "--channels", "1",
  "--encoding", "signed-integer", "--bits", "16", "--type", "raw", "-",
]);
microphone.stdout.on("data", (chunk) => session.write(chunk));
microphone.stdout.on("end", () => session.close());

console.log("Speak, then press Ctrl+C...");
process.on("SIGINT", () => microphone.kill());

const result = await session.result();
console.log(result.text);

Run it, say a sentence, stop the recording. You’ll see the transcript:

Hi, I'm calling about my Best Buy order...

The result also carries per-word confidence, the overall confidence, the audio duration, and a session_id to quote if you contact support. Microphone audio is raw PCM with no header, which is why the config names the sample rate and channel count.

Where to next

Transcribe a short audio file: audio you already have in full, plus every request option: model, language, word timestamps, raw PCM, and calling the API without an SDK.
Transcribe live audio: hand transcribe_live() any source that produces audio over time, such as a call, a WebSocket, or an upload from a browser.
Connection pre-warming: call warm() the moment you know audio is coming to take the handshake off the critical path.
API reference: every header, part, and field.

Need help?

Contact support@assemblyai.com or open a support ticket. Include the session_id from the response.

Getting started

Features

API reference

Advanced

Before you begin

Transcribe from a microphone

Where to next

Need help?

​Before you begin

​Transcribe from a microphone

​Where to next

​Need help?

Before you begin

Transcribe from a microphone

Where to next

Need help?