Documentation

JavaScript
import { LiveTranscription } from '@vocalstack/js-sdk';

const sdk = new LiveTranscription({ apiKey: 'YOUR-API-KEY' });

const stream = await sdk.connect({
  // Optional: Integrate this stream with a Polyglot session
  polyglot_id: 'YOUR-POLYGLOT-SESSION-ID',
  // Optional: language of the speech spoken
  // (this can be used to improve the transcription accuracy)
  language: 'en',
  // Optional: Translate the transcription to these languages
  translations: ['de'],
  // Optional: Stop the stream after this many seconds of inactivity
  timeout_period_s: 60,
  // Optional: Hard stop the stream after this many seconds
  max_duration_s: 300,
});

// Start the stream
stream.start();

// Get audio data from a microphone and send it to the stream
// stream.sendBuffer(buffer);
// *** This is a placeholder for the actual implementation ***

// Manually stop the stream (in this example, after 60 seconds)
// If max_duration_s is set, stopping the stream is optional
setTimeout(() => stream.stop(), 60000);

// Listen for stream transcription data
stream.onData((response) => {
  const { status, data } = response;
  console.log(status); // 'waiting', 'processing', 'done', 'stopping' or 'error'
  if (data) {
    console.log(data.timeline); // an object with the transcription timeline
  }
  if (status === 'done') {
    console.log(data.summary); // a summary of the transcription
    console.log(data.keywords); // an array of keywords
    console.log(data.paragraphs); // the entire transcription in paragraph form
  }
});

获取音频流数据的方式将根据您要执行转录操作的环境而有所不同。下面是一些例子,说明如何做到这一点:

服务器上

在 NextJS 中,您应该安装一个可以从您的设备中获取音频数据的软件包,然后将其转发到 VocalStack API。下面是一个例子:

JavaScript
const mic = require('mic');

// Create a new instance of the microphone utility
const micInstance = mic();

// Get the audio input stream
const micStream = micInstance.getAudioStream();

// Capture the audio data from the microphone
micStream.on('data', (data) => {
  stream.sendBuffer(data); // send the buffer data to the VocalStack API
});

// Start capturing audio from the microphone
micInstance.start();

在网络浏览器上

在网络浏览器中,您可能想要使用. 媒體錄音機 如下例所示,API 是可以使用的。 (使用等软件包也是一个好主意。 记录 改善浏览器兼容性)

JavaScript
// Request access to the microphone
const mediaStream = await navigator.mediaDevices.getUserMedia({ audio: true });

// Create a MediaRecorder instance to capture audio data
const mediaRecorder = new MediaRecorder(mediaStream);

// Event handler to process audio data packets
mediaRecorder.ondataavailable = async (event) => {
  const blob = event.data; // this is the audio packet (Blob)
  const buffer = await blob.arrayBuffer(); // convert the Blob to a Buffer
  stream.sendBuffer(buffer); // send the buffer data to the VocalStack API
};

// Start capturing audio, and send it to the stream every second
mediaRecorder.start(1000);

请注意,要在 Web 客户端上访问 VocalStack API,您需要使用授权令牌:

客户端身份验证令牌

为客户端请求创建临时身份验证令牌。在 Web 浏览器中安全实现 API 请求,而不会暴露您的 API 密钥。.

从 HLS 直播流转录

VocalStack API 可用于转录任何 HLS LiveStream URL,包括 Youtube Live、Facebook Live 和 Twitch 等来源。请注意,流 URL 必须是. .m3u8文件扩展名 文件扩展名,表示有效的 HLS (HTTP 直播流媒体)播放列表文件。.

JavaScript
import { LiveTranscription } from '@vocalstack/js-sdk';

const sdk = new LiveTranscription({ apiKey: 'YOUR-API-KEY' });

const stream = await sdk.connect({
  // must be a valid HLS streaming protocol
  livestream_url:
    'http://a.files.bbci.co.uk/media/live/manifesto/audio/simulcast/hls/nonuk/sbr_low/ak/bbc_world_service.m3u8',

  // The rest of these options are the same as for microphone live transcriptons
});

stream.start();

stream.onData((response) => {
  // The response object is the same as the one
  // returned by microphone transcriptions
});