Normalize the some part of text (number, currency, date, etc) to its spoken form for more consistent speech synthesis (sometimes TTS models might read unnormalized text wrong).

For example, before starting audio generation, it will convert

Call my number 2137112342 on Jul 5th, 2024 for the $24.12 payment

to

Call my number two one three seven one one two three four two on july fifth, twenty twenty four for the twenty four dollars twelve cents payment

Note that this feature adds a bit of latency (~100ms) to the whole process.

Language setting

Currently speech normalization is supported for the following languages:

  • English
  • Spanish
  • French
  • German

Other languages will result in a no-op (text not modified) for this step.

If you selected a language that’s not multilingual, it will use that language code to normalize the text (e.g. 1 will be normalized to one if using English). If you selected multilingual, it will auto detect the language based on the generated text and normalize it accordingly.