For voice agents, I'd be interested in how a sentence ends when the user interrupts. That's where the handoff between listening and speaking becomes very noticeable.
The 90 ms streaming start is the part I'd notice first in a live agent — that's where a voice either feels present or laggy. I'm curious how TTS-1 handles the other end of a long response: does pronunciation and pacing stay consistent across a multi-sentence answer, especially with names, numbers, and addresses? And is there a way to pin down how a specific business name or product term should be pronounced without reworking the whole prompt each time?
Replies
Excited for our first text-to-speech model!!
For voice agents, I'd be interested in how a sentence ends when the user interrupts. That's where the handoff between listening and speaking becomes very noticeable.
The 90 ms streaming start is the part I'd notice first in a live agent — that's where a voice either feels present or laggy. I'm curious how TTS-1 handles the other end of a long response: does pronunciation and pacing stay consistent across a multi-sentence answer, especially with names, numbers, and addresses? And is there a way to pin down how a specific business name or product term should be pronounced without reworking the whole prompt each time?