custom-using.md

copyright	lastupdated	subcollection
years 2015, 2025	2025-02-18	text-to-speech

{{site.data.keyword.attribute-definition-list}}

Using a custom model for speech synthesis

{: #custom-using}

Once you create a custom model and populate it with custom entries, you use it by passing its customization ID (GUID) with the customization_id query parameter of the HTTP GET or POST /v1/synthesize method or the WebSocket /v1/synthesize method. When you include a customization ID, you must call a synthesize method with credentials for the instance of the service that owns the specified custom model. {: shortdesc}

Examples of using a custom model

{: #custom-using-examples}

The first two examples generate a custom pronunciation for IEEE that is based on entries from the indicated custom model. The custom pronunciation is used instead of the default pronunciation from the service's regular pronunciation rules.

• The HTTP GET /v1/synthesize method:

  [IBM Cloud]{: tag-ibm-cloud}

  curl -X GET -u "apikey:{apikey}" \
  --header "Accept: audio/flac" \
  --output ieee.flac \
  "{url}/v1/synthesize?text=IEEE&customization_id={customization_id}"

  {: pre}

  [IBM Cloud Pak for Data]{: tag-cp4d} [IBM Software Hub]{: tag-teal}

  curl -X GET \
  --header "Authorization: Bearer {token}" \
  --header "Accept: audio/flac" \
  --output ieee.flac \
  "{url}/v1/synthesize?text=IEEE&customization_id={customization_id}"

  {: pre}

• The HTTP POST /v1/synthesize method:

  [IBM Cloud]{: tag-ibm-cloud}

  curl -X POST -u "apikey:{apikey}" \
  --header "Content-Type: application/json" \
  --header "Accept: audio/flac" \
  --data "{\"text\":\"IEEE\"}" \
  --output ieee.flac \
  "{url}/v1/synthesize?customization_id={customization_id}"

  {: pre}

  [IBM Cloud Pak for Data]{: tag-cp4d} [IBM Software Hub]{: tag-teal}

  curl -X POST \
  --header "Authorization: Bearer {token}" \
  --header "Content-Type: application/json" \
  --header "Accept: audio/flac" \
  --data "{\"text\":\"IEEE\"}" \
  --output ieee.flac \
  "{url}/v1/synthesize?customization_id={customization_id}"

  {: pre}

The third example establishes a WebSocket connection with the /v1/synthesize method. The request uses the indicated custom model to synthesize text that is passed over the connection.

var access_token = '{access_token}';
var wsURI = '{ws_url}/v1/synthesize'
  + '?access_token=' + access_token
  + '&voice=en-US_AllisonV3Voice'
  + '&customization_id=={customization_id}';
var websocket = new WebSocket(wsURI);

{: codeblock}