What a rule does
Each rule maps a source word or phrase to the text you want the voice to say. Rules are applied before chunking, so the generated sections receive the same prepared spelling wherever the term appears.
This is a text-preparation control. It is not a guarantee of a provider-specific phoneme or SSML interpretation, and it does not replace listening to a short test sample.
- Add product names, people, places, and acronyms that recur across projects.
- Prefer a readable replacement over a complicated phonetic spelling.
- Test the rule in the same voice tier used for the finished read.
Keep the dictionary production-safe
Use one canonical spelling per term and document why the replacement exists. If the spoken form changes, regenerate the sections that contain the term so the audio and SRT reflect the new project revision.