Listing information about models

The IBM Watson® Speech to Text service provides methods for listing information about all of its available models or about a specific large speech model, previous- or next-generation model.

Model information

Regardless of whether you list information about all available models or about a specific model, the service returns the same output for a model:

name is the name of the model that you use in a request.
language is the language identifier of the model (for example, en-US).
rate identifies the minimum acceptable sampling rate in Hertz for audio that is used with the model.
url is the URI for the model.
description provides a brief description of the model.
supported_features describes the additional service features that are supported with the model:
- custom_language_model is a boolean that indicates whether you can create custom language models that are based on the model.
- IBM Cloud custom_acoustic_model is a boolean that indicates whether you can create custom acoustic models that are based on the model.
- low_latency is a boolean that indicates whether you can use the low_latency parameter with a next-generation model. The service includes this field only for next-generation models. Large speech models and previous-generation models do not support the low_latency parameter.
- speaker_labels indicates whether you can use the speaker_labels parameter with the model.
The speaker_labels field returns true for all models. However, speaker labels are supported as beta functionality for only a limited number of languages. For more information, see Speaker labels.

Listing all models

You use the HTTP GET /v1/models method to list information about all available models. The service returns an array of JSON objects that provides information about all of the models.

The order in which the service returns models can change from call to call. So, do not rely on an alphabetized or static list of models. Because the models are returned as an array of JSON objects, the order has no bearing on programmatic uses of the response.

List all models example

The following example lists all models that are supported by the service:

IBM Cloud

curl -X GET -u "apikey:{apikey}" \
"{url}/v1/models"

IBM Cloud Pak for Data

curl -X GET \
--header "Authorization: Bearer {token}" \
"{url}/v1/models"

The response is abbreviated to show only the first few models.

{
  "models": [
    {
      "name": "pt-BR_NarrowbandModel",
      "language": "pt-BR",
      "url": "{url}/v1/models/pt-BR_NarrowbandModel",
      "rate": 8000,
      "supported_features": {
        "custom_language_model": true,
        "custom_acoustic_model": true,
        "speaker_labels": true
      },
      "description": "Brazilian Portuguese narrowband model."
    },
    {
      "name": "ko-KR_BroadbandModel",
      "language": "ko-KR",
      "url": "{url}/v1/models/ko-KR_BroadbandModel",
      "rate": 16000,
      "supported_features": {
        "custom_language_model": true,
        "custom_acoustic_model": true,
        "speaker_labels": true
      },
      "description": "Korean broadband model."
    },
    {
      "name": "fr-FR_BroadbandModel",
      "language": "fr-FR",
      "url": "{url}/v1/models/fr-FR_BroadbandModel",
      "rate": 16000,
      "supported_features": {
        "custom_language_model": true,
        "custom_acoustic_model": true,
        "speaker_labels": true
      },
      "description": "French broadband model."
    },
    . . .
  ]
}

Listing a specific model

You use the HTTP GET /v1/models/{model_id} method to list information about a specified model. The service returns information only for that model.

List a specific previous-generation model example

The following example shows information about the previous-generation US English broadband model, en-US_BroadbandModel:

IBM Cloud

curl -X GET -u "apikey:{apikey}" \
"{url}/v1/models/en-US_BroadbandModel"

IBM Cloud Pak for Data

curl -X GET \
--header "Authorization: Bearer {token}" \
"{url}/v1/models/en-US_BroadbandModel"

The model supports language model customization, acoustic model customization, and speakers labels.

{
  "rate": 16000,
  "name": "en-US_BroadbandModel",
  "language": "en-US",
  "url": "{url}/v1/models/en-US_BroadbandModel",
  "supported_features": {
    "custom_language_model": true,
    "custom_acoustic_model": true,
    "speaker_labels": true
  },
  "description": "US English broadband model."
}

List a specific next-generation model example

The following example shows information about the next-generation US English telephony model, en-US_Telephony:

IBM Cloud

curl -X GET -u "apikey:{apikey}" \
"{url}/v1/models/en-US_Telephony"

IBM Cloud Pak for Data

curl -X GET \
--header "Authorization: Bearer {token}" \
"{url}/v1/models/en-US_Telephony"

The model supports low latency, speakers labels, and language model customization. It does not support acoustic model customization.

{
   "name": "en-US_Telephony",
   "rate": 8000,
   "language": "en-US",
   "description": "US English telephony model for narrowband audio (8kHz)",
   "supported_features": {
      "custom_language_model": true,
      "custom_acoustic_model": false,
      "low_latency": true,
      "speaker_labels": true
   },
   "url": "{url}/v1/models/en-US_Telephony"
}

List a specific large speech model example

The following example shows information about the English large speech model, en-US:

IBM Cloud

curl -X GET -u "apikey:{apikey}" \
"{url}/v1/models/en-US"

IBM Cloud Pak for Data

curl -X GET \
--header "Authorization: Bearer {token}" \
"{url}/v1/models/en-US"

The model supports speakers labels, and language model customization. It does not support acoustic model customization and low latency.

{
  "name": "en-US",
  "rate": 8000,
  "language": "en-US",
  "description": "Large US English model",
  "supported_features": {
    "custom_language_model": true,
    "custom_acoustic_model": false,
    "low_latency": false,
    "speaker_labels": true
    },
  "url": "{url}/v1/models/en-US"
}