Check and correct pronunciation

Names, place names, abbreviations, and invented terms often need a pronunciation decision before recording. Agree on the intended reading and test it in context.

For AI voices

1. Choose a voice that supports the script's language.
2. Generate a short sample containing the difficult word.
3. If the selected voice offers a pronunciation control, check its supported format before using it.
4. Regenerate the affected passage and listen again.

A pronunciation annotation does not change an already-generated audio file. Updating a pronunciation guide is different from updating the current recording.

If the reading is still incorrect

Test another voice or use a human recording for the passage. Listen to the result in the full sentence rather than checking the word in isolation.

For human performers

Add a concise note giving the agreed pronunciation, especially for recurring names and specialized terms. Provide a spoken reference when needed through the available collaboration channel.

Keep a project pronunciation list so different performers use the same reading. For a word with several meanings, note the intended context instead of applying one pronunciation to every occurrence.

Keep the manuscript accurate

Do not replace a name with a different word merely to obtain a sound unless the change is deliberate and acceptable for the production. Check that directions added to the script will not be read aloud accidentally.

After a pickup or regeneration, listen, update changed-audio transcripts if needed, and prepare a new mix.

See Performance guidance, Preview audio, and Upload recordings.