Custom pronunciations not working with Chirp3 HD voices

Gemini API
,
1

I’m trying to use custom pronunciations with Google Cloud TTS following the exact example from the Chirp3 HD documentation:

{

"input": {

  "text": "There is a dog in the boat",

  "custom_pronunciations": {

    "phrase": "dog",

    "phonetic_encoding": "PHONETIC_ENCODING_X_SAMPA",

    "pronunciation": "\\"k{t"

  }

},

"voice": {

  "language_code": "en-US",

  "name": "en-us-Chirp3-HD-Leda"

},

"audio_config": {

  "audio_encoding": "LINEAR16"

}

}

The API call succeeds and returns audio, but the word “dog” is still pronounced as “dog” instead of “cat” (k{t in X-SAMPA).

Questions:

1. Do Chirp3 HD voices actually support custom pronunciations?

2. Is there something wrong with the request format?

3. Are there any undocumented limitations?

The documentation shows this exact example but it doesn’t work in practice. Any help would be appreciated.

2

{
“input”: {
“text”: “There is a dog in the boat”
},
“custom_pronunciations”: [
{
“phrase”: “dog”,
“phonetic_encoding”: “PHONETIC_ENCODING_X_SAMPA”,
“pronunciation”: “"k{t”
}
],
“voice”: {
“language_code”: “en-US”,
“name”: “en-us-Chirp3-HD-Leda”
},
“audio_config”: {
“audio_encoding”: “LINEAR16”
}
}

Mungkin ini bisa membuatmu paham

3

Terima kasih banyak, sekarang saya lebih paham.