how to control how big of a chunk is forwarded from llm to tts in livekit python library voice pipeline agent

Thanks for reaching out, Amjad. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

I want to change the size of the text to ensure complete words are sent in hindi language with eleven labs tts

If you want to tweak the text chunk size for Hindi when using Eleven Labs TTS in LiveKit, you can do that easily with the

chunk_length_schedule

parameter. This lets you set chunk sizes in characters so that complete words are sent smoothly. Here’s a quick example:

tts_11labs = elevenlabs.TTS(
    model="eleven_turbo_v2_5",
    voice_id="bIHbv24MWmeRgasZH58o",
    chunk_length_schedule=[80, 120, 200, 260],
    language="hi"
)

Some helpful things to keep in mind: • The

chunk_length_schedule

takes a list of numbers, each one representing how many characters per chunk—for example,

[80, 120, 200, 260]

. • Each chunk size should be between 50 and 500 characters. • Starting with a smaller chunk size can really help with syllable alignment in Hindi, making the speech sound more natural. If you have any questions or want me to walk you through it, just let me know—I’m here to help!