Migrate code from version 2024-11-15 to version 2025-10-15

Use the speech to text REST API for fast transcription, batch transcription, and custom speech. This article describes changes from version 2024-11-15 to version 2025-10-15.

To summarize the changes in this version:

• The Transcribe API has new features enhanced mode, and phrase list.
• The Projects API returns (absent in version 2024-11-15), and has some changes.

Transcription API changes

Request structure

• New endpoint:

  POST <your_endpoint>/speechtotext/transcriptions:transcribe?api-version=2025-10-15
  

• Headers and form data:
  • Content-Type: multipart/form-data
  • Ocp-Apim-Subscription-Key: $KEY
  • Form fields: definition, audio

Example:

curl --request POST \
  --url '<your_endpoint>/speechtotext/transcriptions:transcribe?api-version=2025-10-15' \
  --header 'Content-Type: multipart/form-data' \
  --header 'Ocp-Apim-Subscription-Key: $KEY' \
  --form 'definition=$DEFINITION' \
  --form 'audio=@C:\workspace\audios\test.wav'

Definition object updates

• Removed:
  • "models" dictionary (no longer in request definition)
• Added:
  • "disfluencyRemoval" (boolean): Removes filler words (such as "um" and "uh")
  • "phraseList": Now supports biasingWeight for recognition bias tuning
  • "enhancedMode" object includes:
    • enabled (boolean)
    • task (such as "translate")
    • targetLanguage (such as "ko")
    • prompt (array of instructions or lexical boosts)

Example:

{
  "locales": ["en-US"],
  "profanityFilterMode": "Masked",
  "diarization": {
    "enabled": true,
    "maxSpeakers": 6
  },
  "channels": [0],
  "disfluencyRemoval": true,
  "enhancedMode": {
    "enabled": true,
    "task": "translate",
    "targetLanguage": "ko",
    "prompt": [
      "Provide lexical output",
      "Boost the terms: CONTOSO, AAZZ; Replace ‘50cents’ to ’50-Cents’"
    ]
  },
  "phraseList": {
    "phrases": ["Kenichi Kumatani", "John McDonough", "Bhiksha Raj"],
    "biasingWeight": 1.6
  }
}

Result structure

• Channel-based output:
  • Results are organized per channel
• Phrase segmentation:
  • Each phrase includes channel, start and end time, speaker, text, and word-level confidence

Projects API changes

New features

• Foundry project name:
  • New property: foundryProjectName in Create, Get, Update, List APIs
• Project creation:
  • Projects are created through Azure Resource Manager (ARM) conventions
  • locale is now required for custom speech projects

Example:

POST {endpoint}/speechtotext/projects?api-version=2025-10-15
Headers:
  Ocp-Apim-Subscription-Key: <YOUR_SUBSCRIPTION_KEY>
  Content-Type: application/json
Body:
{
  "locale": "en-US",
  "displayName": "My speech project",
  "foundryProjectName": "MyFoundrySpeechProject"
}

Project listing and filtering

• Filter by Foundry project name:

  GET {endpoint}/speechtotext/projects?filter=foundryProjectName eq 'MyFoundrySpeechProject'&api-version=2025-10-15
  

Next steps

• Speech to text REST API
• Speech to text REST API 2025-10-15 reference documentation