Voice
On the Voice tab, choose the voice provider and a voice. Native agents offer ElevenLabs (using an ElevenLabs key connected under Integrations) or the platform's native audio voices. Pick the TTS model, then fine-tune pitch and speaking rate.
Browse and preview voices under Resources → Voice Library, filter by gender, mark favourites, and apply a voice to an agent from there.
Pronunciation
For brand, street and product names the voice gets wrong, add a pronunciation: the word and its phonemes in IPA or CMU notation.
Speech-to-text
The Speech-to-Text card sets the transcription provider and model, and takes recognition hints: names and terms the caller is likely to say.
Language model
The LLM Model card picks the model provider and model, with temperature and max tokens.