@langwatch/scenario
    Preparing search index...

    Variable ELEVENLABS_TTS_MODELConst

    ELEVENLABS_TTS_MODEL: "eleven_v3"

    Default ElevenLabs TTS model. eleven_v3 is the only model that honors inline paralinguistic markers like [shouting], [laughs]. The SDK's default (eleven_multilingual_v2) reads them as text — see Python's tts.py:107 for the angry-customer regression that drove the hardcode.