Why is my TTS voice mispronouncing acronyms and product names?

TTS systems often mispronounce acronyms and product names because they don’t know the intended pronunciation from the text alone.

Common reasons:

• Acronyms can be ambiguous
  TTS may read them letter-by-letter, as a word, or in a way that depends on context.
  Example: “NASA” vs “SQL” vs “GIF”.

• Product names often don’t follow normal spelling rules
  Brand names may be stylized, intentionally unusual, or borrowed from another language.
  Example: “Xiaomi,” “Figma,” “Kafka,” “NVIDIA.”

• The TTS engine uses text normalization rules
  It may apply generic pronunciation rules that work for ordinary words but not special terms.

• No pronunciation dictionary entry exists
  If the system doesn’t have a custom lexicon or pronunciation model for the term, it guesses.

• Context is insufficient
  Some models need surrounding words to infer whether something is a name, acronym, or abbreviation.

What you can do:

1. Add phonetic spellings or pronunciation overrides
   Many TTS platforms support SSML, phoneme tags, or custom dictionaries.

2. Rewrite the text slightly
   For example, expand acronyms the first time:

   • “AWS” → “Amazon Web Services (AWS)”

3. Use SSML if supported
   You can specify pronunciation more directly with <phoneme> or <say-as>.

4. Check capitalization and punctuation
   Sometimes formatting changes how the TTS interprets a term.

5. Train or configure a custom pronunciation lexicon
   Especially helpful for brands, internal product names, and technical terms.

If you want, I can help you fix a specific acronym or product name and suggest the exact pronunciation markup for your TTS engine.