Issue

The text-to-speech (TTS) system incorrectly pronounced formatting characters (e.g., asterisks * and double quotes ") as words, resulting in unnatural speech output. For example, asterisks were pronounced as ‘aster,’ and brand names like fastmcp were mispronounced.

Culprit

The system passed raw text, including formatting characters and lowercase brand names, directly to the TTS engine without preprocessing. The TTS system interpreted formatting characters as literal words to be spoken, rather than as visual markers.

Solution

A preprocessing step was introduced to clean text before TTS conversion by removing formatting characters and correcting brand names. This ensures proper pronunciation while maintaining the original text for display and alignment purposes.