WEBVTT

00:00:00.000 --> 00:00:01.256
Thanks for making the time.

00:00:01.336 --> 00:00:03.680
I want to start with the part you said was
messy.

00:00:03.760 --> 00:00:07.640
Right, so the messy part is that the file
never arrives the way you expect it.

00:00:07.720 --> 00:00:09.040
Someone records on a phone,

00:00:09.120 --> 00:00:12.100
someone else on a laptop, and the levels
are nothing alike.

00:00:12.180 --> 00:00:14.150
We stopped sending the audio anywhere.

00:00:14.230 --> 00:00:16.523
Once the whole thing ran on the machine
that

00:00:16.603 --> 00:00:19.004
recorded it, half the problems were just
gone.

00:00:19.084 --> 00:00:21.539
What surprised me was how much of the time
went

00:00:21.619 --> 00:00:23.750
into the download, not the transcription.

00:00:23.830 --> 00:00:27.580
The model is about 200 megabytes, and
after the first run it is cached.

00:00:27.660 --> 00:00:30.600
On the laptop an hour of audio takes about
three minutes.

00:00:30.680 --> 00:00:33.780
On the desktop with a proper graphics
card, well under one.

00:00:33.860 --> 00:00:36.280
That was the moment it stopped feeling
like a demo.

00:00:36.360 --> 00:00:37.800
Let me add a pause here.

00:00:37.880 --> 00:00:41.160
And then continue with the second
paragraph about subtitle formats.

00:00:41.240 --> 00:00:43.560
SRT works on every video platform.

00:00:43.640 --> 00:00:45.800
VTT is what browsers use.

00:00:45.880 --> 00:00:47.880
plain text is 4 notes.