WEBVTT

00:00:00.000 --> 00:00:01.920
Thanks for making the time.

00:00:02.000 --> 00:00:05.920
I want to start with the part you said was
messy.

00:00:06.000 --> 00:00:11.920
Right, so the messy part is that the file
never arrives the way you expect it.

00:00:12.000 --> 00:00:14.118
Someone records on a phone,

00:00:14.198 --> 00:00:18.920
someone else on a laptop, and the levels
are nothing alike.

00:00:19.000 --> 00:00:21.920
We stopped sending the audio anywhere.

00:00:22.000 --> 00:00:25.440
Once the whole thing ran on the machine
that

00:00:25.520 --> 00:00:29.120
recorded it, half the problems were just
gone.

00:00:29.200 --> 00:00:32.538
What surprised me was how much of the time
went

00:00:32.618 --> 00:00:35.520
into the download, not the transcription.

00:00:35.600 --> 00:00:41.400
The model is about 200 megabytes, and
after the first run it is cached.

00:00:41.480 --> 00:00:45.673
On the laptop an hour of audio takes about
three minutes.

00:00:45.753 --> 00:00:50.097
On the desktop with a proper graphics
card, well under one.

00:00:50.177 --> 00:00:53.920
That was the moment it stopped feeling
like a demo.

00:00:54.000 --> 00:00:55.808
Let me add a pause here.

00:00:55.888 --> 00:01:01.000
And then continue with a second paragraph
about subtitle formats.

00:01:01.163 --> 00:01:05.323
SRT works on every video platform.

00:01:05.403 --> 00:01:08.523
VTT is what browsers use.

00:01:08.603 --> 00:01:10.603
Plain text is for notes.