WEBVTT

00:00:01.605 --> 00:00:02.485
<v Speaker 1>Um alright.

00:00:04.965 --> 00:00:08.405
<v Speaker 1>Aryan come in and say something and uh let's see if this thing works.

00:00:09.825 --> 00:00:12.625
<v Speaker 2>Hi, I don't know if you work. I don't think you work.

00:00:13.043 --> 00:00:17.203
<v Speaker 1>Alright, so under the hood this is using Pia Note, whatever, which on the

00:00:17.203 --> 00:00:23.140
<v Speaker 1>diarization benchmark had um I think uh the best NVIDIA world class was 14%,

00:00:23.620 --> 00:00:25.940
<v Speaker 1>and this had 20% error rate.

00:00:25.785 --> 00:00:28.665
<v Speaker 3>What if I'm speaking start speaking in the world?

00:00:29.245 --> 00:00:33.965
<v Speaker 4>Well we can see if it identifies a new speaker that came in recently.