Azure AI Speech

API

Azure speech API that scores pronunciation, fluency and prosody in 33 locales

Fiyat: Free F0 tier, then pay per audio hour
Erişim: Resource key or Entra ID token

Hakkında

Microsoft's speech service, now Azure Speech in Foundry Tools. Pronunciation assessment scores read-aloud or free speech for accuracy, fluency, completeness and prosody down to the phoneme, in 33 locales (no Turkish; prosody in en-US only). Billed as speech to text; prosody costs extra.

Neler yapabilirsin

Score a learner reading a sentence aloud, down to each word, syllable and phoneme
Rate the fluency and prosody of free speech in a language app (prosody in en-US only)
Assess pronunciation in 33 locales such as English, Spanish and Chinese, but not Turkish

Başlarken

Create a Speech or Foundry resource in the Azure portal (F0 is free)
Copy the resource key and name
Send up to 30 s of WAV over REST, or use the Speech SDK for longer audio

Örnek kod

PA=$(printf '%s' '{"ReferenceText":"Good morning.","GradingSystem":"HundredMark","Granularity":"Phoneme","Dimension":"Comprehensive"}' | base64 | tr -d '\n')

curl -X POST "https://$SPEECH_RESOURCE.cognitiveservices.azure.com/stt/speech/recognition/conversation/cognitiveservices/v1?language=en-US&format=detailed" \
  -H "Ocp-Apim-Subscription-Key: $SPEECH_KEY" \
  -H "Content-Type: audio/wav; codecs=audio/pcm; samplerate=16000" \
  -H "Accept: application/json" \
  -H "Pronunciation-Assessment: $PA" \
  --data-binary @good-morning.wav

Ayrıntılar

Barındırma: Hazır hizmet (bulut)
Kullanılabildiği yerler: Tüm dünya
Resmi SDK'lar: C#, Java, Python, JavaScript/TypeScript, Swift
MCP sunucusu: Yok
Entegrasyonlar: Azure