Pronunciation Assessment API
Send a recording and the text the speaker was meant to say; get back how well each word and each phoneme was pronounced. The API runs Azure AI Speech pronunciation assessment with phoneme granularity, prosody and miscue detection turned on, so you do not have to manage an Azure subscription, audio conversion or long recordings yourself.
Read the API documentation →Three calls
Ask for an upload URL, upload the WAV file to it, then ask for the assessment. Every call carries your key in the X-API-Key header. The audio is deleted once it has been scored.
# 1. Ask for an upload URL
curl -X POST https://platform.vocametrix.com/api/get-blob-url \
-H "X-API-Key: $VOCAMETRIX_API_KEY"
# {"uploadURL": "...", "blobURL": "..."}
# 2. Upload the recording
curl -X PUT "$UPLOAD_URL" -H "x-ms-blob-type: BlockBlob" --data-binary @reading.wav
# 3. Score it against the reference text
curl -X POST https://platform.vocametrix.com/api/pronunciation-assessment \
-H "X-API-Key: $VOCAMETRIX_API_KEY" -H "Content-Type: application/json" \
-d "{\"blobURL\": \"$BLOB_URL\", \"referenceText\": \"The weather is nice today\", \"locale\": \"en-US\"}"What comes back
Scores from 0 to 100 for accuracy (how close the phonemes are to a native speaker), fluency (pauses between words), completeness (how much of the reference text was said) and an overall pronunciation score, with the same detail for each word and each phoneme. Each word also carries an error type: omission, insertion, mispronunciation, unexpected break, missing break or monotone. The response follows the Azure JSON format.
Languages and recording length
The locale field takes any language supported by Azure pronunciation assessment, for example en-US, en-GB, fr-FR, es-ES, de-DE or pt-BR. The prosody score (stress, intonation, rhythm) exists only for en-US. Recordings longer than 30 seconds are scored in segments; the response then gives the overall scores and a score per word, without phoneme detail.
Pricing
You pay for the seconds of audio you send, with no rounding up. Credits come in one-time packs, not a subscription: Developer Pack 40min at 5 EUR and Developer Pack 400min at 15 EUR. Creating an account is free, but there is no free tier on the API: the first call already needs a pack.
SDKs
Official clients wrap the same calls: npm install vocametrix for JavaScript and TypeScript, pip install vocametrix for Python.
Also measure the voice itself
The same key opens the acoustic voice measures: AVQI, CPP, jitter, shimmer, HNR and more.
Frequently asked questions
Is there a free tier?
No. The account is free, and the website comes with a 7-day trial, but API calls are paid from a credit pack from the first one.
Where do I find my API key?
On your profile page once you are signed in to Vocametrix.
Is the recording stored?
No. The uploaded file is deleted once the assessment has been returned.