如何缩短Twilio语音识别Gather动作的响应延迟？

使用Twilio和Node.js开发语音识别系统，目标是在通话中用户停止说话后立即播放音频片段，但目前存在严重延迟：用户停止说话后，Twilio调用Gather动作URL处理语音识别结果的间隔约为4-5秒。该延迟已记录在Twilio日志中（），严重影响应用响应速度。

预期行为：用户停止说话后立即触发动作URL。

已尝试的操作：

• 调整speechTimeout：使用speechTimeout: 'auto'及其他参数值，尝试快速检测语音结束
• 分析日志：确认延迟并非来自服务器端处理，而是Twilio内部问题
• 查阅文档：查看Twilio Gather文档寻找优化方案
• 参考同类问题：查阅相关问题但未找到直接解决延迟的方案

app.post('/voice-response', (req, res) => {
  logWithTimestamp('Handling voice response');

  const twiml = new twilio.twiml.VoiceResponse();

  // Gather speech input with a timeout
  twiml.gather({
    input: ['speech'],
    action: `${process.env.BASE_URL}/gather`, // This URL handles the gathered speech
    speechTimeout: '1', // Automatically detect when speech has ended
    // enhanced: true,
    speechModel: "experimental_conversations",
  });

  // If no input is received, play a message
  twiml.say('You need to say something');

  res.type('text/xml');
  res.send(twiml.toString());
  logWithTimestamp('Sent TwiML response for gathering speech');
});

// Endpoint to handle the gather action response
app.post('/gather', (req, res) => {
  logWithTimestamp(`Incoming Gather Request: ${JSON.stringify(req.body)}`);

  const twiml = new twilio.twiml.VoiceResponse();

  if (req.body.SpeechResult) {
    logWithTimestamp(`Speech detected: ${req.body.SpeechResult}`);
    // If speech is detected, acknowledge the speech
    twiml.play(`${process.env.BASE_URL}/voice/output.mp3`);
  } else {
    logWithTimestamp('No speech detected, playing default message');
    // If no speech is detected, play an audio message
    twiml.say('You need to say something');
  }

  res.type('text/xml');
  res.send(twiml.toString());
  logWithTimestamp('Sent TwiML response for gather action');
});

内容的提问来源于stack exchange，提问作者Huzaifa Jalil