#how can we handle proper word spelling ? also phone numbers usually are weirdly spoken (both in spa

1 messages · Page 1 of 1 (latest)

crisp mothBOT
#
AI-generated Answer for xenaroth_ih

Short answer:

  • Proper word/name pronunciation (Spanish): Use Pronunciation Dictionaries. In Studio, add rules in the Pronunciations Editor. For Spanish, use Alias rules (text substitutions). SSML phoneme tags are only supported on English models (Eleven English v1, Flash v2, Turbo v2). (elevenlabs.io)
  • Phone numbers in Spanish: Either write numbers out in words in the script, or enable text normalization and set the language. Via API, use apply_text_normalization='on' and language_code='es'. Normalization can add latency. On the website normalization is on by default; in Studio it’s auto by default (can be set to always on in Project settings > Advanced). (help.elevenlabs.io)

Implementation details (API):

  • Attach dictionaries when converting TTS: pass pronunciation_dictionary_locators to POST /v1/text-to-speech/:voice_id. Create dictionaries by uploading a PLS file, then reference its id and version_id. (elevenlabs.io)
  • Managing dictionaries programmatically: see the “Using pronunciation dictionaries” guide (notes that phoneme tags work only for English; use Alias for other languages, including Spanish). (elevenlabs.io)
  • List existing dictionaries: GET /v1/pronunciation-dictionaries. (elevenlabs.io)

Practical guidance:

  • For Spanish brand names/terms, add Alias rules in a project dictionary (Studio > Pronunciations Editor) to enforce a consistent reading. (help.elevenlabs.io)
  • For acronyms or tricky spellings, you can also alter the input text (e.g., phonetic re-spellings) if you’re not using dictionaries. (help.elevenlabs.io)
  • For phone numbers, when not using normalization, write them out explicitly in Spanish words to remove ambiguity. (help.elevenlabs.io)

References:

  • Why numbers/acronyms may be misread + normalization and language_code guidance. (help.elevenlabs.io)
  • TTS API parameters (apply_text_normalization, language_code, pronunciation_dictionary_locators). (elevenlabs.io)
  • Pronunciations Editor (Studio) and dictionary behavior. (help.elevenlabs.io)
  • Phoneme support scope (models; English only). (help.elevenlabs.io)
  • Using pronunciation dictionaries (programmatic; English-only phonemes; Alias for other languages). (elevenlabs.io)
  • Create/list pronunciation dictionaries via API. (elevenlabs.io)
  • Forcing pronunciation via text tweaks (when not using dictionaries).
graceful igloo
#

this is for both api for tts and agents throguh console. the thing is that we cant use dictionaries, since users phones and emails will vary a lot