Tutors that speak 40 languages, and listen too.

Read every lesson aloud, voice two-person dialogues, and run speaking practice with an AI partner. Iodine speaks the line, transcribes the learner’s answer and speaks the reply, all from one balance.

Hear it

  • 40 languages for lessons and dialogue
  • $0.009 an hour to transcribe learners
  • {slow} slower speech for beginners

How it works

  1. 01

    Speak the lesson

    Send each line with its language code. Wrap new words in {slow} for beginners.

  2. 02

    Hear the learner

    Send their recording to speech to text with the same language code.

  3. 03

    Answer back

    Have your LLM reply and speak it, or stream it over the realtime socket.

import os, requests

H = {"Authorization": f"Bearer {os.environ['IODINE_API_KEY']}"}

# 1. The tutor asks, in Spanish, with the new word slowed down
r = requests.post("https://api.iodine.dev/v1/tts", headers=H, json={
    "text": "¿Qué quieres {slow}desayunar{/slow} hoy?",
    "voice": "mia", "model": "v2", "language": "es", "stream": False,
})
open("question.wav", "wb").write(r.content)

# 2. The learner answers; check what they said
with open("answer.wav", "rb") as f:
    t = requests.post("https://api.iodine.dev/v1/stt", headers=H,
                      files={"file": f}, data={"language": "es"}).json()
print(t["text"])  # "Quiero café con tostadas"

What it costs

50,000 learners a month, each hearing 30 minutes of lessons and speaking for 10.

1,500,000 minutes of lessons on Iodine $11,250.00

500,000 minutes of answers transcribed $75.00

Total for the month $11,325.00

The same speech at ElevenLabs v4 Turbo's published price $60,000.00

At Iodine’s pay-as-you-go prices. Other providers at their published pay-as-you-go rates; see the price comparison.

Questions

English, Chinese, Spanish, Hindi, Arabic, French, Portuguese, Bengali, Russian, Japanese, German, Korean, Urdu, Indonesian, Turkish, Vietnamese, Italian, Tamil, Telugu, Punjabi, Thai, Tagalog, Persian, Polish, Ukrainian, Dutch, Malay, Romanian, Swahili, Greek, Czech, Hungarian, Swedish, Hebrew, Danish, Finnish, Norwegian, Slovak, Bulgarian and Croatian.

Yes. Every v2 voice speaks all of them, so a tutor character can keep one voice across every course you offer.

Speech to text tells you what the learner said, so you can compare it with what they meant to say. It doesn’t score individual sounds.

Start building with Iodine.

Top up from $5. No plan, seats or minimums.

Get an API key   ›
iodine