LiveKit does not have a built-in pronunciation dictionary for Text-to-Speech (TTS). However, it supports customization of word pronunciations during real-time audio streaming through the use of Speech Synthesis Markup Language (SSML). This includes tags like

phoneme

for phonetic pronunciation and

lexicon

for custom dictionaries defining specific word pronunciations. Additionally, LiveKit offers a

before_tts_cb

callback in the

VoicePipelineAgent

class that lets you modify the text before sending it to the TTS engine, allowing you to replace words with preferred pronunciations. Certain TTS providers integrated with LiveKit, like ElevenLabs, also support custom pronunciations via SSML. So while LiveKit does not have a built-in dictionary, these mechanisms allow you to achieve customized and accurate pronunciations in your TTS streams. Sources: Agent speech | LiveKit Docs | VoicePipelineAgent | LiveKit Docs | VoicePipelineAgent | LiveKit Docs | ElevenLabs TTS integration guide | LiveKit Docs | Agent speech and audio | LiveKit Docs