WEBVTT
Kind: captions
Language: en

00:00:00.000 --> 00:00:01.870 align:start position:0%
 
Hey<00:00:00.400><c> everyone.</c>

00:00:01.870 --> 00:00:01.880 align:start position:0%
Hey everyone.
 

00:00:01.880 --> 00:00:03.710 align:start position:0%
Hey everyone.
Imagine<00:00:02.480><c> you're</c><00:00:02.600><c> in</c><00:00:02.720><c> a</c><00:00:02.760><c> massive</c><00:00:03.200><c> meeting</c><00:00:03.560><c> with</c>

00:00:03.710 --> 00:00:03.720 align:start position:0%
Imagine you're in a massive meeting with
 

00:00:03.720 --> 00:00:05.390 align:start position:0%
Imagine you're in a massive meeting with
20<00:00:04.040><c> people.</c>

00:00:05.390 --> 00:00:05.400 align:start position:0%
20 people.
 

00:00:05.400 --> 00:00:07.670 align:start position:0%
20 people.
Everyone<00:00:05.920><c> is</c><00:00:06.040><c> talking</c><00:00:06.440><c> at</c><00:00:06.560><c> once</c><00:00:07.240><c> and</c><00:00:07.400><c> you</c><00:00:07.480><c> need</c>

00:00:07.670 --> 00:00:07.680 align:start position:0%
Everyone is talking at once and you need
 

00:00:07.680 --> 00:00:09.830 align:start position:0%
Everyone is talking at once and you need
a<00:00:07.760><c> perfect</c><00:00:08.200><c> transcript</c><00:00:08.840><c> that</c><00:00:09.000><c> knows</c><00:00:09.280><c> exactly</c>

00:00:09.830 --> 00:00:09.840 align:start position:0%
a perfect transcript that knows exactly
 

00:00:09.840 --> 00:00:12.750 align:start position:0%
a perfect transcript that knows exactly
who<00:00:10.000><c> said</c><00:00:10.280><c> what</c><00:00:10.800><c> the</c><00:00:10.920><c> moment</c><00:00:11.240><c> they</c><00:00:11.400><c> say</c><00:00:11.640><c> it.</c>

00:00:12.750 --> 00:00:12.760 align:start position:0%
who said what the moment they say it.
 

00:00:12.760 --> 00:00:15.790 align:start position:0%
who said what the moment they say it.
Usually,<00:00:13.520><c> that's</c><00:00:13.760><c> expensive</c><00:00:14.360><c> and</c><00:00:14.520><c> slow.</c>

00:00:15.790 --> 00:00:15.800 align:start position:0%
Usually, that's expensive and slow.
 

00:00:15.800 --> 00:00:19.070 align:start position:0%
Usually, that's expensive and slow.
But<00:00:16.080><c> today,</c><00:00:16.640><c> Meta</c><00:00:16.960><c> just</c><00:00:17.280><c> changed</c><00:00:17.680><c> the</c><00:00:17.800><c> game.</c>

00:00:19.070 --> 00:00:19.080 align:start position:0%
But today, Meta just changed the game.
 

00:00:19.080 --> 00:00:20.990 align:start position:0%
But today, Meta just changed the game.
Meta<00:00:19.480><c> Superintelligence</c><00:00:20.480><c> Labs</c><00:00:20.840><c> has</c>

00:00:20.990 --> 00:00:21.000 align:start position:0%
Meta Superintelligence Labs has
 

00:00:21.000 --> 00:00:22.510 align:start position:0%
Meta Superintelligence Labs has
officially<00:00:21.520><c> launched</c><00:00:21.920><c> News</c><00:00:22.240><c> Voice</c>

00:00:22.510 --> 00:00:22.520 align:start position:0%
officially launched News Voice
 

00:00:22.520 --> 00:00:24.110 align:start position:0%
officially launched News Voice
Transcribe.

00:00:24.110 --> 00:00:24.120 align:start position:0%
Transcribe.
 

00:00:24.120 --> 00:00:26.150 align:start position:0%
Transcribe.
This<00:00:24.480><c> is</c><00:00:24.600><c> a</c><00:00:24.640><c> brand</c><00:00:25.000><c> new</c><00:00:25.240><c> audio</c><00:00:25.600><c> perception</c>

00:00:26.150 --> 00:00:26.160 align:start position:0%
This is a brand new audio perception
 

00:00:26.160 --> 00:00:27.510 align:start position:0%
This is a brand new audio perception
model<00:00:26.520><c> that</c><00:00:26.720><c> handles</c><00:00:27.120><c> streaming</c>

00:00:27.510 --> 00:00:27.520 align:start position:0%
model that handles streaming
 

00:00:27.520 --> 00:00:29.710 align:start position:0%
model that handles streaming
transcription<00:00:28.320><c> and</c><00:00:28.520><c> speaker</c><00:00:28.880><c> diarization</c>

00:00:29.710 --> 00:00:29.720 align:start position:0%
transcription and speaker diarization
 

00:00:29.720 --> 00:00:32.749 align:start position:0%
transcription and speaker diarization
for<00:00:29.920><c> over</c><00:00:30.160><c> 20</c><00:00:30.560><c> people</c><00:00:30.920><c> simultaneously.</c>

00:00:32.749 --> 00:00:32.759 align:start position:0%
for over 20 people simultaneously.
 

00:00:32.759 --> 00:00:34.870 align:start position:0%
for over 20 people simultaneously.
The<00:00:32.960><c> most</c><00:00:33.320><c> shocking</c><00:00:33.680><c> part?</c>

00:00:34.870 --> 00:00:34.880 align:start position:0%
The most shocking part?
 

00:00:34.880 --> 00:00:36.510 align:start position:0%
The most shocking part?
The<00:00:35.120><c> price.</c>

00:00:36.510 --> 00:00:36.520 align:start position:0%
The price.
 

00:00:36.520 --> 00:00:39.070 align:start position:0%
The price.
Meta<00:00:37.000><c> is</c><00:00:37.200><c> offering</c><00:00:37.600><c> this</c><00:00:37.880><c> for</c><00:00:38.000><c> just</c><00:00:38.320><c> 18</c><00:00:38.760><c> cents</c>

00:00:39.070 --> 00:00:39.080 align:start position:0%
Meta is offering this for just 18 cents
 

00:00:39.080 --> 00:00:41.550 align:start position:0%
Meta is offering this for just 18 cents
per<00:00:39.280><c> hour</c><00:00:39.520><c> of</c><00:00:39.680><c> processed</c><00:00:40.240><c> audio.</c>

00:00:41.550 --> 00:00:41.560 align:start position:0%
per hour of processed audio.
 

00:00:41.560 --> 00:00:43.830 align:start position:0%
per hour of processed audio.
To<00:00:41.800><c> put</c><00:00:42.000><c> that</c><00:00:42.200><c> in</c><00:00:42.320><c> perspective,</c><00:00:43.200><c> competitors</c>

00:00:43.830 --> 00:00:43.840 align:start position:0%
To put that in perspective, competitors
 

00:00:43.840 --> 00:00:46.470 align:start position:0%
To put that in perspective, competitors
like<00:00:44.040><c> AssemblyAI</c><00:00:44.920><c> or</c><00:00:45.120><c> Deepgram</c><00:00:45.720><c> can</c><00:00:45.920><c> cost</c><00:00:46.360><c> two</c>

00:00:46.470 --> 00:00:46.480 align:start position:0%
like AssemblyAI or Deepgram can cost two
 

00:00:46.480 --> 00:00:48.230 align:start position:0%
like AssemblyAI or Deepgram can cost two
to<00:00:46.600><c> three</c><00:00:46.880><c> times</c><00:00:47.280><c> more</c><00:00:47.600><c> when</c><00:00:47.760><c> you</c><00:00:47.880><c> add</c><00:00:48.120><c> in</c>

00:00:48.230 --> 00:00:48.240 align:start position:0%
to three times more when you add in
 

00:00:48.240 --> 00:00:50.470 align:start position:0%
to three times more when you add in
speaker<00:00:48.680><c> identification.</c>

00:00:50.470 --> 00:00:50.480 align:start position:0%
speaker identification.
 

00:00:50.480 --> 00:00:53.590 align:start position:0%
speaker identification.
Muse<00:00:50.880><c> isn't</c><00:00:51.200><c> just</c><00:00:51.480><c> cheap,</c><00:00:52.040><c> it's</c><00:00:52.280><c> smart.</c>

00:00:53.590 --> 00:00:53.600 align:start position:0%
Muse isn't just cheap, it's smart.
 

00:00:53.600 --> 00:00:55.430 align:start position:0%
Muse isn't just cheap, it's smart.
It<00:00:53.880><c> processes</c><00:00:54.520><c> audio</c><00:00:54.920><c> in</c><00:00:55.040><c> tiny</c>

00:00:55.430 --> 00:00:55.440 align:start position:0%
It processes audio in tiny
 

00:00:55.440 --> 00:00:57.790 align:start position:0%
It processes audio in tiny
80-millisecond<00:00:56.240><c> chunks</c><00:00:57.000><c> using</c><00:00:57.320><c> a</c><00:00:57.400><c> trick</c>

00:00:57.790 --> 00:00:57.800 align:start position:0%
80-millisecond chunks using a trick
 

00:00:57.800 --> 00:00:59.910 align:start position:0%
80-millisecond chunks using a trick
called<00:00:58.080><c> adaptive</c><00:00:58.560><c> delay.</c>

00:00:59.910 --> 00:00:59.920 align:start position:0%
called adaptive delay.
 

00:00:59.920 --> 00:01:01.950 align:start position:0%
called adaptive delay.
It<00:01:00.120><c> can</c><00:01:00.280><c> wait</c><00:01:00.520><c> a</c><00:01:00.600><c> split</c><00:01:00.920><c> second</c><00:01:01.320><c> if</c><00:01:01.440><c> the</c><00:01:01.560><c> speech</c>

00:01:01.950 --> 00:01:01.960 align:start position:0%
It can wait a split second if the speech
 

00:01:01.960 --> 00:01:03.870 align:start position:0%
It can wait a split second if the speech
is<00:01:02.080><c> messy</c><00:01:02.480><c> or</c><00:01:02.640><c> commit</c><00:01:03.000><c> immediately</c><00:01:03.560><c> if</c><00:01:03.680><c> it's</c>

00:01:03.870 --> 00:01:03.880 align:start position:0%
is messy or commit immediately if it's
 

00:01:03.880 --> 00:01:05.150 align:start position:0%
is messy or commit immediately if it's
clear.

00:01:05.150 --> 00:01:05.160 align:start position:0%
clear.
 

00:01:05.160 --> 00:01:07.270 align:start position:0%
clear.
It<00:01:05.440><c> supports</c><00:01:05.920><c> over</c><00:01:06.120><c> 70</c><00:01:06.560><c> languages</c><00:01:07.160><c> and</c>

00:01:07.270 --> 00:01:07.280 align:start position:0%
It supports over 70 languages and
 

00:01:07.280 --> 00:01:09.350 align:start position:0%
It supports over 70 languages and
handles<00:01:07.680><c> things</c><00:01:08.000><c> like</c><00:01:08.200><c> code-switching</c><00:01:09.200><c> where</c>

00:01:09.350 --> 00:01:09.360 align:start position:0%
handles things like code-switching where
 

00:01:09.360 --> 00:01:11.350 align:start position:0%
handles things like code-switching where
people<00:01:09.720><c> jump</c><00:01:10.040><c> between</c><00:01:10.400><c> languages</c><00:01:11.000><c> in</c><00:01:11.120><c> one</c>

00:01:11.350 --> 00:01:11.360 align:start position:0%
people jump between languages in one
 

00:01:11.360 --> 00:01:12.830 align:start position:0%
people jump between languages in one
sentence.

00:01:12.830 --> 00:01:12.840 align:start position:0%
sentence.
 

00:01:12.840 --> 00:01:15.390 align:start position:0%
sentence.
Now,<00:01:13.480><c> is</c><00:01:13.600><c> it</c><00:01:13.720><c> a</c><00:01:13.760><c> world</c><00:01:14.120><c> record?</c>

00:01:15.390 --> 00:01:15.400 align:start position:0%
Now, is it a world record?
 

00:01:15.400 --> 00:01:17.110 align:start position:0%
Now, is it a world record?
Not<00:01:15.760><c> quite.</c>

00:01:17.110 --> 00:01:17.120 align:start position:0%
Not quite.
 

00:01:17.120 --> 00:01:19.470 align:start position:0%
Not quite.
Speechmatics<00:01:18.040><c> can</c><00:01:18.240><c> handle</c><00:01:18.600><c> up</c><00:01:18.760><c> to</c><00:01:18.920><c> 100</c>

00:01:19.470 --> 00:01:19.480 align:start position:0%
Speechmatics can handle up to 100
 

00:01:19.480 --> 00:01:22.070 align:start position:0%
Speechmatics can handle up to 100
speakers<00:01:20.360><c> and</c><00:01:20.560><c> Amazon</c><00:01:21.040><c> Transcribe</c><00:01:21.720><c> handles</c>

00:01:22.070 --> 00:01:22.080 align:start position:0%
speakers and Amazon Transcribe handles
 

00:01:22.080 --> 00:01:23.350 align:start position:0%
speakers and Amazon Transcribe handles
30.

00:01:23.350 --> 00:01:23.360 align:start position:0%
30.
 

00:01:23.360 --> 00:01:25.310 align:start position:0%
30.
But<00:01:23.640><c> for</c><00:01:23.760><c> most</c><00:01:24.080><c> businesses</c><00:01:24.640><c> building</c><00:01:25.000><c> meeting</c>

00:01:25.310 --> 00:01:25.320 align:start position:0%
But for most businesses building meeting
 

00:01:25.320 --> 00:01:27.710 align:start position:0%
But for most businesses building meeting
assistants<00:01:26.000><c> or</c><00:01:26.120><c> call</c><00:01:26.440><c> analytics,</c><00:01:27.360><c> 20</c>

00:01:27.710 --> 00:01:27.720 align:start position:0%
assistants or call analytics, 20
 

00:01:27.720 --> 00:01:29.750 align:start position:0%
assistants or call analytics, 20
speakers<00:01:28.200><c> is</c><00:01:28.360><c> plenty.</c>

00:01:29.750 --> 00:01:29.760 align:start position:0%
speakers is plenty.
 

00:01:29.760 --> 00:01:31.990 align:start position:0%
speakers is plenty.
Plus,<00:01:30.440><c> Meta</c><00:01:30.760><c> is</c><00:01:30.920><c> topping</c><00:01:31.320><c> the</c><00:01:31.440><c> charts</c><00:01:31.800><c> for</c>

00:01:31.990 --> 00:01:32.000 align:start position:0%
Plus, Meta is topping the charts for
 

00:01:32.000 --> 00:01:33.550 align:start position:0%
Plus, Meta is topping the charts for
accuracy.

00:01:33.550 --> 00:01:33.560 align:start position:0%
accuracy.
 

00:01:33.560 --> 00:01:35.550 align:start position:0%
accuracy.
In<00:01:33.800><c> recent</c><00:01:34.200><c> benchmarks</c><00:01:34.840><c> from</c><00:01:35.040><c> Artificial</c>

00:01:35.550 --> 00:01:35.560 align:start position:0%
In recent benchmarks from Artificial
 

00:01:35.560 --> 00:01:38.790 align:start position:0%
In recent benchmarks from Artificial
Analysis,<00:01:36.520><c> Muse</c><00:01:36.880><c> hit</c><00:01:37.040><c> a</c><00:01:37.120><c> 3.1%</c><00:01:38.280><c> word</c><00:01:38.560><c> error</c>

00:01:38.790 --> 00:01:38.800 align:start position:0%
Analysis, Muse hit a 3.1% word error
 

00:01:38.800 --> 00:01:41.430 align:start position:0%
Analysis, Muse hit a 3.1% word error
rate<00:01:39.440><c> beating</c><00:01:39.760><c> out</c><00:01:39.920><c> rivals</c><00:01:40.320><c> like</c><00:01:40.560><c> OpenAI,</c>

00:01:41.430 --> 00:01:41.440 align:start position:0%
rate beating out rivals like OpenAI,
 

00:01:41.440 --> 00:01:43.870 align:start position:0%
rate beating out rivals like OpenAI,
Google,<00:01:42.000><c> and</c><00:01:42.160><c> ElevenLabs.</c>

00:01:43.870 --> 00:01:43.880 align:start position:0%
Google, and ElevenLabs.
 

00:01:43.880 --> 00:01:45.790 align:start position:0%
Google, and ElevenLabs.
It's<00:01:44.200><c> also</c><00:01:44.560><c> leading</c><00:01:44.880><c> in</c><00:01:45.000><c> diarization</c>

00:01:45.790 --> 00:01:45.800 align:start position:0%
It's also leading in diarization
 

00:01:45.800 --> 00:01:48.270 align:start position:0%
It's also leading in diarization
accuracy,<00:01:46.760><c> which</c><00:01:46.960><c> means</c><00:01:47.280><c> fewer</c><00:01:47.600><c> mistakes</c><00:01:48.120><c> in</c>

00:01:48.270 --> 00:01:48.280 align:start position:0%
accuracy, which means fewer mistakes in
 

00:01:48.280 --> 00:01:50.590 align:start position:0%
accuracy, which means fewer mistakes in
figuring<00:01:48.760><c> out</c><00:01:48.960><c> who</c><00:01:49.120><c> is</c><00:01:49.240><c> talking.</c>

00:01:50.590 --> 00:01:50.600 align:start position:0%
figuring out who is talking.
 

00:01:50.600 --> 00:01:52.630 align:start position:0%
figuring out who is talking.
While<00:01:50.920><c> there</c><00:01:51.080><c> are</c><00:01:51.200><c> some</c><00:01:51.440><c> tradeoffs</c><00:01:52.200><c> like</c><00:01:52.440><c> no</c>

00:01:52.630 --> 00:01:52.640 align:start position:0%
While there are some tradeoffs like no
 

00:01:52.640 --> 00:01:54.870 align:start position:0%
While there are some tradeoffs like no
word-level<00:01:53.200><c> timestamps</c><00:01:53.880><c> yet,</c><00:01:54.400><c> the</c><00:01:54.520><c> value</c>

00:01:54.870 --> 00:01:54.880 align:start position:0%
word-level timestamps yet, the value
 

00:01:54.880 --> 00:01:56.710 align:start position:0%
word-level timestamps yet, the value
here<00:01:55.120><c> is</c><00:01:55.240><c> incredible.</c>

00:01:56.710 --> 00:01:56.720 align:start position:0%
here is incredible.
 

00:01:56.720 --> 00:01:58.950 align:start position:0%
here is incredible.
For<00:01:57.000><c> enterprise</c><00:01:57.560><c> developers,</c><00:01:58.560><c> this</c><00:01:58.760><c> is</c><00:01:58.880><c> a</c>

00:01:58.950 --> 00:01:58.960 align:start position:0%
For enterprise developers, this is a
 

00:01:58.960 --> 00:02:00.950 align:start position:0%
For enterprise developers, this is a
massive<00:01:59.400><c> win</c><00:01:59.680><c> for</c><00:01:59.880><c> building</c><00:02:00.280><c> live</c><00:02:00.640><c> AI</c>

00:02:00.950 --> 00:02:00.960 align:start position:0%
massive win for building live AI
 

00:02:00.960 --> 00:02:02.950 align:start position:0%
massive win for building live AI
assistants<00:02:01.680><c> and</c><00:02:01.880><c> meeting</c><00:02:02.200><c> tools</c><00:02:02.600><c> without</c>

00:02:02.950 --> 00:02:02.960 align:start position:0%
assistants and meeting tools without
 

00:02:02.960 --> 00:02:04.790 align:start position:0%
assistants and meeting tools without
breaking<00:02:03.320><c> the</c><00:02:03.440><c> bank.</c>

00:02:04.790 --> 00:02:04.800 align:start position:0%
breaking the bank.
 

00:02:04.800 --> 00:02:06.910 align:start position:0%
breaking the bank.
Meta<00:02:05.160><c> is</c><00:02:05.320><c> making</c><00:02:05.680><c> high-end</c><00:02:06.120><c> voice</c><00:02:06.479><c> AI</c>

00:02:06.910 --> 00:02:06.920 align:start position:0%
Meta is making high-end voice AI
 

00:02:06.920 --> 00:02:09.150 align:start position:0%
Meta is making high-end voice AI
accessible<00:02:07.520><c> to</c><00:02:07.680><c> everyone.</c>

00:02:09.150 --> 00:02:09.160 align:start position:0%
accessible to everyone.
 

00:02:09.160 --> 00:02:10.950 align:start position:0%
accessible to everyone.
What<00:02:09.360><c> do</c><00:02:09.479><c> you</c><00:02:09.600><c> think?</c>

00:02:10.950 --> 00:02:10.960 align:start position:0%
What do you think?
 

00:02:10.960 --> 00:02:13.430 align:start position:0%
What do you think?
Is<00:02:11.200><c> 18</c><00:02:11.680><c> cents</c><00:02:12.000><c> an</c><00:02:12.160><c> hour</c><00:02:12.400><c> low</c><00:02:12.600><c> enough</c><00:02:12.959><c> to</c><00:02:13.080><c> change</c>

00:02:13.430 --> 00:02:13.440 align:start position:0%
Is 18 cents an hour low enough to change
 

00:02:13.440 --> 00:02:16.030 align:start position:0%
Is 18 cents an hour low enough to change
how<00:02:13.560><c> your</c><00:02:13.720><c> company</c><00:02:14.160><c> uses</c><00:02:14.560><c> AI?</c>

00:02:16.030 --> 00:02:16.040 align:start position:0%
how your company uses AI?
 

00:02:16.040 --> 00:02:18.270 align:start position:0%
how your company uses AI?
Let<00:02:16.240><c> me</c><00:02:16.360><c> know</c><00:02:16.560><c> in</c><00:02:16.640><c> the</c><00:02:16.720><c> comments.</c>

00:02:18.270 --> 00:02:18.280 align:start position:0%
Let me know in the comments.
 

00:02:18.280 --> 00:02:19.710 align:start position:0%
Let me know in the comments.
Like,<00:02:18.560><c> subscribe,</c><00:02:19.160><c> and</c><00:02:19.320><c> turn</c><00:02:19.600><c> on</c>

00:02:19.710 --> 00:02:19.720 align:start position:0%
Like, subscribe, and turn on
 

00:02:19.720 --> 00:02:22.720 align:start position:0%
Like, subscribe, and turn on
notifications.