Support for audioEncoding as a number suddenly dropped

[Masadow]

Hi, we realized that using name in voice (which is fine according to typescript typings) is no longer working and has been dropped yesterday.

It broke our production environment without prior notice. Changing it to ssmlGender works but does not produce the same voice.

We also changed audioEncoding from number to string as a safety measure as it's no longer documented.

Old request :

const request = {
  request: {
    audioConfig: {
      audioEncoding: 2
    },
    input: {
      ssml: '<speak><par><media xml:id="backgroundAudio" end="text0.end+3.0s" ><audio src="https://ector-v3-dev.s3-eu-west-1.amazonaws.com/music/corporate-achievement.mp3" soundLevel="-12dB" /></media><media xml:id="text0">blahblah</media></par></speak>'
    },
    voice: {
      languageCode: 'fr-FR',
      name: 'fr-FR-Wavenet-A'
    }
  }
}

New request :

const request = {
  request: {
    audioConfig: {
      audioEncoding: 'MP3'
    },
    input: {
      ssml: '<speak><par><media xml:id="backgroundAudio" end="text0.end+3.0s" ><audio src="https://ector-v3-dev.s3-eu-west-1.amazonaws.com/music/corporate-achievement.mp3" soundLevel="-12dB" /></media><media xml:id="text0">blahblah</media></par></speak>'
    },
    voice: {
      languageCode: 'fr-FR',
      ssmlGender: 'FEMALE'
    }
  }
}

[yil532]

Yes, thanks for the notice. For more information, here is the documentation and code sample in Node: https://cloud.google.com/text-to-speech/docs/libraries#client-libraries-usage-nodejs/. The key points you mentioned have been included here.