Vapi is the Voice AI platform for developers.

Vapi AI

Hey guys, I have a question: Does the TTS start as soon as it receives the first chunk of text or does it wait until the whole message is complete? Also, how does the TTS build words? For example, if the final completion is 'Hello' and it receives 'H' at the start, does the TTS read it right away or does it wait for the next letter?

Or maybe TTS is waiting for the words? I just want to know how it works

What happens in the case of Custom LLMs? My initial impression is that Vapi for its clusters starts TTS as soon as the first tokens are received, but is it the same for a custom LLM?

I bet the AI helper button in the corner in the HELP docs will help

And watch the calls logs, you can see the "Chunking" https://share.zight.com/OAuZy8v5

<@665414293830762506> model output is sent to TTS as sentence such as "hello how are you?" then after this TTS output is streamed out.