Skip to main content
The Wavix Speech Analytics API is available exclusively for Flex Pro users.
The Wavix Speech Analytics API supports two ways to transcribe phone calls:
  1. Transcribe an active phone call that was initiated or received through Wavix.
  2. Transcribe a phone call recorded outside Wavix by uploading an audio file. For instructions on uploading and transcribing external recordings, see Call transcription (file upload).
This guide describes how to transcribe calls that went through Wavix and how to search those transcriptions for keywords and phrases. To search call transcriptions for keywords or phrases, first enable Call recording and Call transcription services on your SIP trunks and phone numbers. You can do this in the Wavix app or by using the SIP trunks and DIDs endpoints.

​
Prerequisites

Before you can access the Wavix Speech Analytics API, sign up for Wavix.

​
Create a Wavix account

  1. Sign up for a Wavix account using your business email address.
  2. Confirm your email address and phone number during the sign-up process.
  3. Wait for your account to be approved by the Wavix team.
  4. After approval, choose either the Wavix Flex or Flex Pro account level.
Once your account is approved, you can access the Wavix app and APIs.

​
Find your API key

Wavix uses API keys to authenticate requests. To find API keys associated with your account:
  1. Sign in to your Wavix account.
  2. Go to Administration → API Keys.
  3. Copy the API key you want to use, or create a new one by clicking Create new.

​
How Wavix call transcription works

When you activate transcription on a SIP trunk or phone number, Wavix automatically converts recorded calls into plain text. Most calls are transcribed within 10 minutes, depending on call length and queue size. After transcription, Wavix detects the call language and labels speakers as agent or client based on call direction.

​
Request transcription for a specific call

To request a transcription for a specific call:
  1. Use the CDR API to find the call’s unique identifier (uuid).
  2. Request the transcription: 
    GET https://api.wavix.com/v1/cdr/uuid/transcription?appid=your_api_key
If successful, you receive an HTTP 200 OK response with the full transcription. The transcription is divided into blocks by speaker, with start and end times in milliseconds from call answer. Sample response: 
{
    "transcript": {
        "19293741161": "Hello",
        "19293741160": "Hi"
    },
    "turns": [
        {
            "phone_number": "19293741161",
            "s": 600,
            "e": 700,
            "text": "Hi"
        }
    ],
    "uuid": "1234",
    "language": "en",
    "duration": 10,
    "charge": "0.003",
    "status": "completed",
    "transcription_date": "2023-01-09T10:04:39.734Z"
}

​
Search for calls containing keywords or phrases

To search for calls with transcription containing specific keywords or phrases, use the following method 
POST https://api.wavix.com/v1/cdr
Response sample: 
{
   "type": "placed",
   "from": "2023-01-01",
   "to": "2023-01-31",
   "transcription": {
       "status": "completed",
       "language": "en",
       "agent": {
           "must": [
               "Hello",
               "Thank you"
           ],
           "match": [
               "Nope",
               "Maybe"
           ],
           "exclude": [
               "Richard",
               "Issue resolved"
           ]
       },
       "client": {
           "must": [
               "Hello",
               "Bye, Larry"
           ],
           "match": [
               "Nope",
               "Maybe"
           ],
           "exclude": [
               "Richard",
               "Issue resolved"
           ]
       },
       "any": {
           "must": [
               "Hello",
               "Bye, Larry"
           ],
           "match": [
               "Nope",
               "Maybe"
           ],
           "exclude": [
               "Richard",
               "Issue resolved"
           ]
       }
   }
}
  • type - use placed for outbound calls or received for inbound calls
  • from - filter results by the lower limit on the date the call was placed or received
  • to - filter results by the upper limit on the date the call was placed or received
  • transcription - filter results by the call transcription parameters:
    • status - filter results by transcription status
    • language - filter results by the language of the call
    • agent - filter results based on keywords and phrases said by a speaker labeled as agent
      • must - find calls with transcription that includes all specified keywords and phrases. The keywords and phrases are combined using logical AND.
      • match - find calls with transcription that includes any specified keyword and phrase. The keywords and phrases are combined using logical OR.
      • exclude - find calls with transcription that do not include any specified keyword and phrase. The keywords and phrases are combined using logical OR.
    • client - filter results based on keywords and phrases said by a speaker labeled as client
      • must - find calls with transcription that includes all specified keywords and phrases. The keywords and phrases are combined using logical AND.
      • match - find calls with transcription that includes any specified keyword and phrase. The keywords and phrases are combined using logical OR.
      • exclude - find calls with transcription that do not include any specified keyword and phrase. The keywords and phrases are combined using logical OR.
    • any - filter results based on keywords and phrases said by any speaker
      • must - find calls with transcription that includes all specified keywords and phrases. The keywords and phrases are combined using logical AND.
      • match - find calls with transcription that includes any specified keyword and phrase. The keywords and phrases are combined using logical OR.
      • exclude - find calls with transcription that do not include any specified keyword and phrase. The keywords and phrases are combined using logical OR.
The type, from, and to are mandatory parameters and cannot be blank. All other parameters are optional. If successful, the method returns the HTTP 200 OK status code. The response will contain a list of call detail records that match the search criteria. Every call detail record contains the unique identifier of the call and a link to the full call transcription. Sample response: 
{
    "items": [
        {
            "uuid": "8891",
            "date": "2022-03-03T13:53:59.000Z",
            "from": "12212123123",
            "to": "18001231233",
            "duration": 72,
            "charge": "0.01",
            "disposition": "ANSWERED",
            "transcription": {
                "uuid": "1234",
                "url": "https://api.wavix.com/v1/cdr/8891/transcription?appid=secret"
            },
            "destination": "United States",
            "sip_trunk": "36465",
            "per_minute": "0.0059"
        }
    ],
    "pagination": {
        "total": 100,
        "total_pages": 10,
        "current_page": 2,
        "per_page": 0
    }
}

​
Transcribe a single call

To transcribe a specific call:
  1. Use the CDR API to find the call’s unique identifier (uuid).
  2. Send a PUT request: 
    PUT https://api.wavix.com/v1/cdr/uuid/retranscribe
  3. Include the following JSON in the request body: 
    {
    "language": "en",
    "webhook_url": "https://your-webhook.url"
}
Parameters:
  • language: (Optional) Force the language for transcription. Wavix uses this instead of auto-detection.
  • webhook_url: (Optional) Webhook URL for transcription status updates.
If successful, you receive an empty HTTP 200 OK response. When transcription is complete, you receive a POST callback to your webhook: 
{
    "uuid": "123",
    "status": "completed"
}
  • uuid: Unique identifier of the recorded call
  • status: completed (success) or failed (error)
⌘I