Elevenlabs TTS plugin doesn't support pronunciation dictionaries

[arvindvs]

Feature Type

Would make my life easier

Feature Description

Currently Elevenlabs supports pronunciation dictionaries with most of their TTS streaming models.

These provide a programmatic way of pronouncing words without any additional prompting required. We use these extensively on other voice AI platforms.

These pronunciation dictionaries are typically defined in the initializeConnection message to Elevenlabs, where the voice settings are also defined.

Elevenlabs reference

I'd like to allow native support of pronunciation dictionaries when instantiating the Elevenlabs plugin in Livekit. This will follow the same structure as VoiceSettings.

Workarounds / Alternatives

Prompting the LLM to provide phonetic-based replacements, which is inconsistent and a waste of context.

Additional Context

No response