Chamelingo API Documentation

The pronunciation scoring endpoint uses a faster-whisper speech-to-text pipeline to evaluate how accurately users pronounce Korean words and phrases.

Score Pronunciation#

Submit an audio recording via multipart form data:

bash

curl -X POST https://api.chamelingo.com/api/v1/pronunciation/score \
  -H "Authorization: Bearer ck_live_YOUR_KEY_HERE" \
  -F "audio=@recording.wav" \
  -F "expected_text=안녕하세요" \
  -F "sensitivity=normal"

Parameters#

Field	Type	Required	Description
`audio`	file	Yes	Audio file (wav, webm, mp3, ogg, max 5MB)
`expected_text`	string	Yes	The text that should have been spoken
`sensitivity`	string	No	`easy`, `normal` (default), `strict`

Response#

JSON

{
  "overall_score": 85,
  "words": [
    {
      "expected": "안녕하세요",
      "recognized": "안녕하세요",
      "probability": 0.92,
      "status": "correct"
    }
  ],
  "full_text": "안녕하세요",
  "expected_text": "안녕하세요",
  "feedback": "Good pronunciation overall.",
  "duration": 2.3
}

Sensitivity Levels#

Level	Description
`easy`	Lenient — good for beginners, accepts approximations
`normal`	Balanced — standard pronunciation expectations
`strict`	Demanding — expects near-native accuracy

Check Service Health#

The pronunciation server runs separately. Check its availability before sending audio:

bash

curl https://api.chamelingo.com/api/v1/pronunciation/health \
  -H "Authorization: Bearer ck_live_YOUR_KEY_HERE"

JSON

{
  "available": true,
  "service": "faster-whisper"
}

Warning

If available is false, the pronunciation server is offline. Score requests will fail until it recovers.

Tips for Best Results#