Prompt
ElevenLabs API custom pronunciation
Latest observation
Yes — ElevenLabs supports custom pronunciation in a few ways, depending on which product you’re using:
1) Pronunciation dictionaries / lexicons
If you want a specific word, name, acronym, or brand to be spoken a certain way, the usual approach is to provide a pronunciation dictionary (sometimes called a lexicon or pronunciation guide) to the TTS request.
Typical use cases:
- Brand names
- Character names
- Acronyms
- Foreign words
You define how the word should be pronounced, and then reference that dictionary when generating speech.
2) Phonetic spelling in the text
A simpler workaround is to write the word phonetically in the input text, for example:
- “NVIDIA” → “en-vid-ee-uh”
- “GIF” → “jif”
This is less precise, but often works.
3) SSML-like control
If you’re using a speech stack that supports SSML or similar text markup, you may be able to influence pronunciation that way. Support depends on the exact ElevenLabs endpoint/product, so check the current docs for the model/endpoint you’re using.
4) Voice settings / model choice
Sometimes pronunciation quality improves by:
- switching to a different voice
- using a different model
- improving punctuation/capitalization in the text
Practical recommendation
If your goal is reliable custom pronunciation at scale, use the official pronunciation dictionary / lexicon support if available in your ElevenLabs plan/API version.
If you want, I can help with either of these:
- show the exact ElevenLabs API example for custom pronunciation, or
- help you create a pronunciation entry for a specific word/name.
Brands