Voice Selection Rules

VoiSona Talk selects the voice library for speech synthesis based on a specific order of priority. Understanding these rules helps you get predictable results.

Priority Order

If you don't specify all parameters (language, voice_name, voice_version), the engine follows this logic:

1. Exact Match: The library that matches all three: language, voice_name, and voice_version.
2. Partial Match: The last library in the installed list that matches both language and voice_name.
3. Recent Use: The most recently used voice library for that specific language.
4. Default Fallback: The first library in the installed list that matches the language.

Recommendation

For production environments, always specify both language and voice_name to ensure your application doesn't change voices unexpectedly when new libraries are installed or the user manually changes the voice in the editor.

// Best Practice
await client.synthesizeAndWait({
  text: 'Predictable voice',
  language: 'ja_JP',
  voice_name: 'tanaka-san_ja_JP'
});