Fix pronunciation in the speech text

Viimati uuendatud 27. okt 2025

Learn how to fix pronunciation by adjusting how specific words sound in generated speech.

Try it in the app
Fix pronunciation in the speech text in a few simple steps.

When using Generate Speech (beta) in Firefly, you might notice that certain words aren't pronounced as expected. These instances could include proper names, technical terms, or words with multiple possible pronunciations. Fix Pronunciation allows you to adjust how these words sound to match your preferences and make sure they are pronounced exactly as intended. 

Before you begin:

Enter speech text or upload a text file, then customize the voice, accent, and other settings.

In the text editor, identify and highlight the words that need pronunciation adjustment.

Select Fix Pronunciation from the context menu.

The text editor window displays a highlighted word, with the Fix Pronunciation button above it.
Use the Fix Pronunciation button to adjust how specific words are spoken in generated speech.

In the text box that appears, type out the word phonetically – as it should sound. For example, type "uhn-fer-GET-uh-buhl" to correctly pronounce "unforgettable".

The pronunciation editor in the Text to speech screen displays a highlighted word and it's corrected phonetic version.
To generate accurate sounds, ensure that the entered text is phonetically correct.

Select the icon to play and preview the adjusted pronunciation.

If you're happy with how the word sounds, select Fix to change just the selected instance.

To apply the same adjusted pronunciation to all instances of this word throughout your audio, select Fix all instead.

Näpunäide:

After fixing the pronunciation, generate a new version and play back the entire audio to ensure the updated pronunciation is applied to the rest of the generated speech.