Build an IVR System with Speeko API

Interactive Voice Response used to sound robotic. In 2026, there's no excuse.

The Modern IVR Stack

• Telephony — Twilio, Vonage, or Plivo for call routing
• Voice generation — Speeko for natural greetings and prompts
• Speech recognition — For caller input interpretation
• Backend logic — Route calls based on caller responses

Sample Twilio Integration

const twilio = require('twilio');
const axios = require('axios');

app.post('/voice', async (req, res) => {
  const twiml = new twilio.twiml.VoiceResponse();

  const audio = await axios.post(
    'https://api.speekoapp.com/v1/tts',
    {
      text: 'Welcome to Acme Corp. Press 1 for sales, 2 for support.',
      voice: 'af_heart',
      format: 'mp3'
    },
    { headers: { Authorization: 'Bearer YOUR_KEY' }, responseType: 'arraybuffer' }
  );

  const audioUrl = await uploadToCDN(audio.data);
  twiml.play(audioUrl);
  twiml.gather({ numDigits: 1, action: '/handle-menu' });

  res.type('text/xml').send(twiml.toString());
});

Caching Strategy

IVR prompts repeat thousands of times. Pre-generate and cache them:

const PROMPTS = {
  welcome: 'ivr/welcome.mp3',
  menu: 'ivr/menu.mp3',
  invalid: 'ivr/invalid.mp3'
};

This drops your TTS costs to near zero for static prompts.

Dynamic Personalization

For personalized messages, generate on-demand:

const greeting = `Hello ${customer.firstName}, your order will arrive tomorrow.`;
const audio = await generateTTS(greeting);

Call Flow Design

• Keep menus short — 4 options max
• Offer "speak to a human" always
• Confirm caller input before acting
• Use natural phrasing, not "press 1 for the first option"
• Respect regional norms (formality, pacing)

Monitoring

Track drop-off at each menu level. If 40% of callers abandon the first menu, redesign it.

Get started with Speeko.