December 25, 2024 • 45 mins
Returning guest, Ian Lavery from Picovice.ai, joins the hosts to talk all things voice recognition. He dives into new languages the company has tackled over the last year (and what languages it plans to tackle next year), how they train their models, and how Picovoice is actually running speech recognition in the browser instead of in the cloud, making things like captioning live streams and real-time chats possible with some of its newer tech Cheetah and Leopard.
He also shares how he wrote a simple podcast transcription app using Picovoice and Express.js, in addition to Picovoice boasting specific SDKs for React, Angular and Vue.
Listen to Ian's first appearance on RRU here where he and the panel went deep into the specifics of voice recognition like security and privacy, understanding it in general, and using it sans big cloud providers.
Sponsors
Picks
Become a supporter of this podcast: https://www.spreaker.com/podcast/react-round-up--6102072/support.
He also shares how he wrote a simple podcast transcription app using Picovoice and Express.js, in addition to Picovoice boasting specific SDKs for React, Angular and Vue.
Listen to Ian's first appearance on RRU here where he and the panel went deep into the specifics of voice recognition like security and privacy, understanding it in general, and using it sans big cloud providers.
Sponsors
- "Waldo, who helps with Android testing"
- Chuck's Resume Template
- Developer Book Club starting with Clean Architecture by Robert C. Martin
- Become a Top 1% Dev with a Top End Devs Membership
Picks
- Ian - Mixpanel: Product Analytics for Mobile, Web, & More
- Paige - Star Trek: Lower Decks - Wikipedia
- TJ - The Great British Bakeoff series
Become a supporter of this podcast: https://www.spreaker.com/podcast/react-round-up--6102072/support.
Listen
Watch
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Welcome to React Roundup, the podcast where we keep you
updated on all things React related. This show is brought
to you by Void and top End Devs. Unvoid provides
high quality design and software development services on a client
friendly business model. Unlike all other software agencies, Unvoid allows
(00:20):
clients to only pay after the work is delivered and approved.
Visit unvoid dot com to learn more and reach out.
If you know a company that needs more professionals to
help with design and software development, that's u n void
dot com and top end Davs helps you stay up
to date with cutting edge technologies like JavaScript, Ruby, Elixir,
(00:44):
and AI. Visit topandevs dot com to join their AIDV
boot camp, weekly community meetups and access expert tutorials. I'm
Lucas Paganini, founder of Onvoid and host of this podcast.
Thank you for tuning in. Let's jump into the episode.
Speaker 2 (01:07):
Hey everybody, and welcome to another episode of React Roundup.
I am your host today TJ Van Tol and with
me on the panel, I have Paige need you House.
Hey everyone, and our special guest today is actually a
React Round of returning Champion. We have even Lovery here. Ian.
Welcome back to the show.
Speaker 3 (01:24):
Hey, thanks for having me back.
Speaker 2 (01:25):
Yeah, so why don't you start, you know, for people,
I think it's show we're looking back? Is the show
is about a year ago. We'll have to look up
the episode number and toss it in the show notes.
But it's been a while, So why don't you tell
people know who you are, what you do in your
background while you're famous, all those sorts of things.
Speaker 3 (01:41):
Yeah. So I work for a speech recognition company called
pegle Boys, and we're a developer focused company that tries
to power developers all over on any platform to have
to bring voice to their platform. So we have a
whole variety of different propus that cover speech to text,
(02:02):
voice activation, wake word, all that, and we just want
everybody to have a voice on their platform. Besides that,
I'm a I do like interactive media hard and I
play bass in a couple of bands.
Speaker 4 (02:16):
That's awesome, not just one band but multiple.
Speaker 3 (02:20):
Yeah, I'm an over at cheap I.
Speaker 2 (02:22):
Guess well cool. So Peka Voice looks interesting. I remember
us talking about it last time, but maybe you can
get an overview of like how it works. Like if
I if I use Peka Voice, what am I? What
am I getting? Am I getting a service that I
can send like audio to you, and it comes back
with the words like what other features maybe you could
give us, like the rundown of everything. It does everything
(02:43):
you do.
Speaker 3 (02:44):
Yeah. So the big thing with us is and our
sort of thing that sets us apart from pretty much
every other voice service is that we're entirely on device
and so there is no there is no service. There's
no cloud API that you're calling to send your audio to,
which I mean, look look around. That's pretty much every
(03:05):
single voice thing is just an API. So we're one
of the only ones out there that is actually giving
you the ability to hold on to your audio data
and your user's audio data and process it on the
device and return Again. We have like a variety of products.
So we have like wakeword detection, where it's just like hey,
(03:25):
Siri and okay Google. It's just all it's doing is
sitting there processing frames of audio, waiting for you to
say the thing, and then when it wakes up, it
does the thing that you tell it to do. But
we also have voice activity detection and which just basically
peaks when it hears somebody talking. And obviously speech to text.
Everyone wants speech to text, so auto transcription of voice.
Speaker 2 (03:49):
Yeah, it's very cool. It's also like one of those
problems that I feel like is it's becoming more commonplace.
We have smart devices in our house. Our phones can
listen to wake words and that sort of thing. But
I still I'm still sort of fascinated by the underlying technology.
Maybe you could just start give us like the world's
simplest rundown of like how does how does it actually
work on the back end? Like do you just have
(04:11):
a whole bunch of like low level C code that
looks for patterns in audio data or like, I don't know,
we don't need it. Sounds like two hours, but I'm.
Speaker 3 (04:20):
Just no, it's a good question. So, I mean, basically,
it's deep learning, right, It's it's it's machine learning. So
we teach through machine learning. We teach a machine a
statistical model of what a word sounds like, or what
a series of sounds sounds like. So we basically take
audio in our actual When we're teaching our machine, all
(04:41):
we're doing is sending it frames of audio that are labeled,
and we get it to remember them and like form
a little statistical pattern, and then it for something like wakeword.
It's just like, hey, remember this pattern of three things.
Just remember that and say, hey, I think I saw it.
So it's a lot more complicated when you get into
speech to text because not only are you teaching it
(05:02):
every sound in the language, but you're also teaching it
every word in the language, because then you're dealing with
audio and writing, which are different things. I think people
think language is a combination of those things, but really
they're two entirely separate things. They're like that there's the
(05:23):
series of sounds you make with your mouth that other
people understand, and then there's the symbols you write them
down with and the grammar and punctuation and everything that
you put into the written form, and they're different, so
we actually have to treat them differently. But you'll see
a lot of the big cloud providers out there. The
reason they got it so right so fast is because
(05:46):
they had such large machines in the cloud in order
to do this, so sort of like it outpaced the
actual progressive voice recognition, and now everything's kind of caught
up and we can actually do it on Devine, which
is a big win because, to be honest, we were
like boiling the ocean for like a while doing speech
(06:07):
to text, and now we can do it on like
a micro controller.
Speaker 4 (06:10):
So if you're using something like Peaco Voice, is it
something that you as a user have to train the
models or the models already there. It's trained. It knows
you're speaking English or it knows you're speaking Spanish, and
it will just it should be smart enough to be
able to take that audio and translate it into the
(06:32):
correct written words.
Speaker 3 (06:34):
Right, So, like for speech to text, for instance, we
basically just have a general language model. You just give it.
We offer eight different languages, and you just give it
the language you want and we'll understand that language. But
we actually use this thing called transfer learning, and we
have a website Peako Voice Console where you can basically
(06:56):
we have sort of a general model, but then you
sort of do train it yourself. Because for something like wakeword,
we have a model that understands a bunch of sounds
in whatever language you give it, but then you want
it to represent a certain series of sounds like okay, Google,
So you literally type that in to our console and
hit train, and then it will pop out a model
(07:18):
that understands that. So that's that's sort of the when
we say you train it, it's not like, oh, you
have to go out and gather four thousand recordings of
this word and you know, submit it to something and
watch statistics go and decide. No. No, it's just like
we are. We did the hard work.
Speaker 2 (07:38):
I was gonna say, because by saying that, you're sort
of implying that you got went out and have four
thousand recordings of these different words, right or like.
Speaker 3 (07:47):
No, No. So the thing is, it's again we've we've
trained the general model, so it understands the sounds we
needed to understand. You just tell us which sounds you
want us you want to form your wake word, and
we pop out a model that's that that just waits
for those series of sounds.
Speaker 2 (08:02):
Interesting because I would have guessed that your building of
the model was to get a bunch of people to
say like it almost seems it kind of breaks my
mind a little bit as possible, right, that you can
sort of general.
Speaker 3 (08:14):
Us the old style the like. So I worked. I
worked at a speech recognition company right out of college.
And what we did we had one of the early
early wakeword engines, and what we would do is we'd
it was all b to be the company. We basically
enter a contract with the company that says, hey, we're
(08:36):
going to go out and gather four thousand recordings of
this wake word, and we're going to train it and
then deliver you the model. And it was very formal,
and that was basically state of the art at the time.
But we're actually a bit past that now because we're
able to use this concept of transfer learning to take
(08:58):
a general model and just kind of pointed in the
right direction. So we no longer need to do all
that all that pounding the pavement asking for people to
say a wake word, because that was a lot of
work and it took months, like every time somebody signed
a contract. And I know because I was running the
crowdsourcing technology for that company, So I would have to
post these jobs and these these people would record it
(09:22):
on their on their like mobile device, and I'd have
to go through all the recordings and like you know,
some people would just yeah. Some people would just you know,
speak their manifesto into the phone, and I'd be like, no, no, no, no.
Speaker 4 (09:37):
So one one thing that I'm curious about is I'm
assuming that when you would do these these wake word gatherings,
you would have to take into account accents, because I
know that that is something that every automated assistant struggles with.
This English accents, Scottish accents, Caribbean accents, all speaking English,
(09:57):
but all slightly differently. So is PEKO voice able to
account for that and be able to interpret, you know,
a deep Southern accent versus maybe a New York Boston accent.
Speaker 3 (10:11):
Yeah, So I mean that that's still a challenge for us.
But I think the reason we're a bit more resilient
to it is because we've trained this general model on
like g'z like ten hundred thousand hours of speech. It's
heard all the accents, not not all the accents, but
it's heard it's heard a lot of variation, so it
(10:34):
tends to be a bit more resilient. When I was
doing the old style where we would get people to record,
that was actually a lot less resilient to it because
we only had like, you know, three hundred participants recording
these wake words, and how much variety are you going
to get between three hundred people? Like? Not enough? But
(10:55):
when we train these general models, we have like tens
of thousands of different speakers, maybe more, so we tend
to be a lot more sensitive to the variations. But
but it is, it is definitely a challenge because even
us as humans, if you hear like a really thick
accent that you're not used to, it can be confusing,
like like we're we're not perfect either with it. So
(11:18):
it's it's it's a challenge.
Speaker 2 (11:20):
So I think you so you added multiple language depart
I believe that's new or at least newish from the
last time we talk. So does that that like more generalizability,
make that easier or I imagine there's still all sorts of
challenges that go into that.
Speaker 3 (11:38):
Yeah, So when you when you actually work with a
totally different language, that's basically starting over because accents is
one thing you've already taught it the series of sounds
in the language, and you're just looking for a combination
of those sounds and those symbols. But when you move
into a new language, there's a new set of symbols,
and there's a new set of sounds. You know, there's
(12:00):
everybody has an inventory. We call it a phonemic inventory,
and it's basically a series of sounds that you hear
in the language, and every language has a different phonemic inventory,
and we need to train the machine to understand only
that inventory of sounds and all the symbols that go
into that. So when we start a new language, we
(12:21):
have to do it completely from scratch. We have to
get new data in that language, We need to get
new text in that language, and we need to do
our best to even understand the language enough to work
with it because we need to listen to these recordings.
We need to normalize the text we get and make
(12:44):
sure it's not like full of symbols and stuff, but
understand it enough so that we actually don't confuse the
machine learning process, and that that could be a real challenge.
It's a lot of work.
Speaker 2 (12:55):
Actually, so it's fascinating. Does that mean like when you
kick off anywe language, I feel like you almost need
to have like a professional linguist on staff for almost
each of these languages, right, Like, or do you like
bring on somebody who's, like, you know, a world class
I don't know, a Spanish linguist to help, or like
(13:15):
like how much of it are you able, like as
a software developer to sort of test on your own
and how much do you have to rely on a
native speaker as the only person that can actually figure
some of these things out.
Speaker 3 (13:26):
Yeah, so we do have like basically our like machine
learning team. They do have to be part linguist, like
because if you've studied languages, you at least understand the components,
and basically every language is just a combination of the components.
So they have a lot of expertise in that field
to understand when they approach a new language how it works.
(13:50):
But then that's not enough. So usually what we'll do
is we'll get somebody, well, we will get a native speaker.
Usually will basically hire somebody on a contract to work
with us to help with the language, because you do
need that expertise. Like, the fact is, even somebody who's
(14:11):
like a language expert, if they sit down to an
entirely new language, they're not going to be able to
understand it enough to do the work that needs to
be done to actually get it to a production ready state.
So we often do need to get a native speaker
in there to provide their input and that will really
(14:31):
speed the process along. We tried to do it without
experts a couple times, and it's just like you just
don't get the performance and you spend a lot more time,
you waste a lot more time. I should say sure.
Speaker 4 (14:44):
I mean that makes a lot of sense. When you
think about getting expertise in anything else, it's a lot.
It will almost undoubtedly go much quicker if you have
somebody who is proficient in whatever it is that you're
trying to do.
Speaker 3 (14:58):
Yeah, well, they can recognize mistakes, grammar and stuff, the
stuff that's really hard to pick up as a non
native speaker.
Speaker 4 (15:04):
Yes, So what languages do you currently offer Peko voice for?
Speaker 3 (15:09):
So we have I believe last year we announced we
had Spanish, French, German, English, and then this year we
added four new languages. We added Japanese, Korean, Portuguese, and Italian.
Speaker 4 (15:26):
These are some tough ones.
Speaker 3 (15:28):
Yeah, well, especially like when you get into the written
forms of Korean and Japanese, they become very challenging. Like
you know, we in English we have twenty six characters.
Japanese has two alphabets of fifty six and then an
additional alphabet of tens of thousands. So yeah, yeah, so
(15:53):
that the text representation of that is really difficult. The
actual spoken version of Japanese is a lot easier than
English because Japanese has fifty six sounds, and they all
map to a combination of characters. English mapping a combination
of characters to the sound is incredibly difficult. Turns out
(16:16):
we made some mistakes early on and we didn't really
fix them.
Speaker 4 (16:20):
I mean, just thinking about the amount of spellings that
we have for the same sounding word based on context,
I can not even imagine how you would be able
to figure that out for a transcript.
Speaker 3 (16:32):
And it's all exceptions in English. It's like, oh, yeah,
it's this unless this or this unless this, and like
here's three different reasons why this rule is wrong.
Speaker 2 (16:41):
And yeah, you see this when you have like younger
kids that are starting to write, and you look at
their writing because they start they don't know the exceptions yet, right,
but they can speak it because they know. So you
get like, they it's words you don't even think about too,
because we internalize them so quickly. Because one of my
kids spelled because ron and then you're like, oh, bill,
(17:02):
because it's pretty easy. But then you think about it
for like half a second and you realize, like, actually,
the word because makes absolutely no sense, like right.
Speaker 3 (17:09):
Like if you try and explain it, you suddenly find
yourself going just is what it is. Yes, just memorize it, yep.
Speaker 4 (17:19):
I mean that's really fantastic that you have taken on
and it sounds like gotten through some very difficult dialects.
Are what are future future languages that you hope to
be able to process as well?
Speaker 3 (17:32):
So we're going, yeah, exactly, So we're going to try
next year. We're going to double our language count again,
I think, and we're going to do going to do Chinese, Vietnamese,
what else? Dutch? I believe, Russian, Polish, I think, yeah,
I can't remember all of them. But you basically need
(17:52):
to be like a fully inclusive speech recognition company, you
basically need like a bare minimum of like fifty languages.
So like we're going to get to like twenty of
the most popular and hold there for a while. Is
kind of our plan because that covers a lot of people,
Like that covers the majority of people, because because even
(18:15):
in the cases where the people might not speak the language,
they usually are like, oh but I speak this what
this more popular language? But to really get up there,
like I mean, you do need to get to like
fifty or something. And I mean Google has like one
hundred and fifty, so you know, it's it's kind of
a never ending thing for us.
Speaker 4 (18:37):
How about Hindi that's a big one.
Speaker 3 (18:39):
Oh yeah, that's actually one of the other ones we're
going to do next year.
Speaker 2 (18:43):
Yeah, so I guess I got to ask one last question.
Are there any languages like you've come to hate, like
because it was like very difficult or.
Speaker 3 (18:56):
It's funny how much you can hate your own language. No, actually,
like seriously, English is the only Like I look at
all other languages we've done, and I'm like, these are
so much easier, like English is. Actually, it's just it
came out of a mess of languages. It was a
lot of combinations that happened over time, and a lot
(19:18):
of them happened during like you know, a lot of
English developed during like illiteracy, and so there's like really
interesting examples you can find of like stuff where it's
just like, oh, yeah, this was just a mistake that happened,
you know, two hundred years ago that they kept in
or actually, I have a fun fact the word dumb.
So you look at that, you're like, why does it
(19:39):
have a be at the end that apparently was there
was a time where the like ruling class of England
was trying to make it harder to write English so
that the peasantry could like pick it up. And they
literally just added some letters to the language here or there,
and we're like, this is the proper way to write it,
(19:59):
and then just to confuse people. And we literally still
have that to this day. So English is so weird.
Speaker 4 (20:06):
So that's why knife has a K in front of it.
Speaker 3 (20:09):
Yeah, yeah, like like stuff like that. I think they
were just messing with us and now we're just like
we have to live with that.
Speaker 2 (20:17):
So I want to pivot a little bit and talk
about the actual web development, like the side where you
might actually use a service like this, because I remember
last time we chat it a little bit too about
common use cases, right, So maybe we could just start
with a review, like how we have a lot of
web developers listen to this show. What do you think, like,
I guess, A, what would using something like this look like?
(20:39):
Like how do you actually get it in an app?
And B I guess like, what are some common use
cases that you see for use on the web as well?
Speaker 3 (20:46):
Right? So one of the big things is obviously on
the web, people are a lot more comfortable calling like
an API and that is what they've come to expect
for speech trek cognition and stuff. But we're actually bringing
the We're actually kind of bringing back the power of
the browser itself. So the I mean the browser is
a virtual environment that can run whatever you want. And
(21:09):
we actually can run entirely in the browser on the
client side. And that's that's big because I mean, in
the these days, we're getting a lot of progressive web apps,
and the sort of web app is a big thing,
especially with like SaaS companies and stuff. So if you're
running like a SaaS company and you you want to
(21:30):
integrate like voice into your console or something, having it
on the client side is is I mean, it lowers
the latency, It gives you a lot more direct control
of what happens when you get boys, and it means
you can be robust to connection issues, which like that
that you know, that's a huge thing. Not everyone has
(21:53):
amazing Internet and you don't want to have to be
I can calls out to an API and just hoping
it comes back for you or feature to work. This
will just work. And also on top of all that,
it's it's less expensive because we're not calling an API,
We're not depending on cloud infrastructure. So you're actually if
(22:13):
you're a developer and you integrate Peka Voice into your
web app, your client is going to be using their
machine to do the processing. So I think it's just
a win win situation for that. Yeah.
Speaker 2 (22:26):
I feel it's especially important considering it's audio too, So
like bandwidth is like you you're not just shipping off
like a couple of things in a query string to
some service. You're like uploading.
Speaker 3 (22:37):
Audio h mega adio.
Speaker 2 (22:40):
Yeah, so the bandwidth consideration is amplified significantly.
Speaker 3 (22:46):
Yeah. No, and it makes everything that that actually allows
for like something like real time audio. Real time audio
is very challenging to do for an API because you
basically need to stream it to the service and have
responses being streamed back. That's that's really expensive. That's like
a constant bandwidth issue. But when you're doing real time
(23:08):
audio and it's all running in your browser on the
client side, it's it's snappy and you can do things
that require timing and.
Speaker 2 (23:16):
Yeah, very cool. And I know I think I remember
from last time too that because one of the ways
you keep it snappy is it's not JavaScript code running
in the browser, right, it's your I don't remember your
exact tech stack, but I know you have some sort
of fancy way of doing that. Maybe you could walk
people through some of the magic and challenges of how
that works.
Speaker 3 (23:35):
Yeah, so our core code is in like C because
we we were trying to keep it as efficient and
snappy as possible. Now C code, and when you think
of C code next to React, you're like, how does
this even work? Like can these two ever talk? But
it turns out they can with WAM, And what we
do is we compile basically all our core code into
(23:58):
a WAHSM binary and then we ship that with our
like MPM package. So when you MPM install Peka Voice,
part of the part of what's going to be shipped
with your website is OURASM blob. And basically wasm's really
cool because it basically just wraps your native code in
(24:21):
JavaScript and then allows you to basically attached to it
like any sort of dynamic library, say, here's the functions
I want to call, here's the data I'm going to
put into it, and then you just call it like
you would any any other library. It's a little trickier
(24:41):
to work with because you're dealing with I mean JavaScript
obviously one of the things we're pretty aware of, and
I'm sure the listeners of your show are aware of.
Is JavaScript is like eh, types whatever, even typescript is
like is like yeah, types, but like you know, a
number is a number, right? Well, see is like what
how many bits is your number? Like he needs to know?
(25:06):
So you start to need to think about that. When
you work with WASM, you start to need to think
about okay, is this a thirty two bit and going
in here? And you need to start to think of okay,
you need to start to think of memory, like okay,
I need to have a pointer. I need to pass
in a pointer here to get something back and then
convert that pointer to like a JavaScript object of some sort.
(25:30):
So it's challenging to work with, but once you get
it working, it's extremely powerful because then we can ship
something that's incredibly complex piece of code and just put
basically a slim interface of JavaScript around it, and then
any jabscript developer can just call is just talking to
(25:51):
it like it's JavaScript. They don't need to worry about
the WAHSM that was our problem. Yeah, So it's challenging
to work with, but I do if anybody's thinking of
has a challenging problem that requires the efficiency of c
Don't be afraid of it. It's not that it's not
that hard, and it is pretty amazing when you start
(26:12):
working with it.
Speaker 4 (26:12):
Actually, Okay, so it works or there there is an
NPM package if you want to use JavaScript with it.
But what if you are a Python developer or maybe
you're working with a micro controler like our do we
know is there are there options for other other languages
like that?
Speaker 3 (26:32):
Yeah, So, I mean we support since we're a developer
focused company, we're pretty obsessed with our SDKs. So I
think for our two most popular products, I think we
have like twenty SDKs for each one, and it covers
all the favorites. And we even have you know, we
(26:53):
have three No, we have four different web SDKs. We
have Vanilla JavaScript, but we also have React, Angular and view.
So it allows we basically wanted to be like, use
it in your favorite environment, like, yeah, use it like
you use anything else in your in your stack, Like
(27:14):
we don't want to just disturb that basically.
Speaker 4 (27:17):
Right, that's awesome. So what are some of the use
cases that you've seen people employing it for recently?
Speaker 3 (27:25):
So we've seen, so we've actually come to some interesting
ones lately. So auto content moderation is a big one
right now. So let's say you're Minecraft or something, or
you're I guess, let's let's go Fortnite, and you have
open audio streams hundreds of thousands of players, and you're
(27:48):
trying to moderate all that. That's that's very difficult. And
it turns out a lot of big companies out there
are using auto moderation, which basically takes that audio and
is basically looking for key phrases let's call them, and
(28:10):
it's just looking to flag them and then and then
they'll usually have you know, a person go in and
inspect the actual content of it and decide whether it
was you know, a mistake or whether it is actually
like a bannable offense. Yeah, so that that's actually a
new exciting one. Also, like call centers, it turns out,
(28:31):
again we've got open phone lines, like like a whole
building full of them, and we're trying to understand, you know,
what's being said on all these different calls, and you
can't have people listening to all that audio. So a
lot of big call center companies need some sort of
automated system to take in all the audio from all
their phones and do something with it. So we're we're
(28:54):
encountering more use cases like that lately.
Speaker 2 (28:57):
Actually, those are both really fascinating. It's funny. The content
moderation one really resonated with me because I play I
don't know. My kids are eleven, so they're right at
that impressionable age, but they're also right in the age
where they want to play like games that are the
sort where they have open audio. So there's a game
we play that's like five y five, so five people
(29:18):
on each team, and it has it has a way
for you to do audio communication. And the very first
thing I did was make sure to shut that off,
like disable it, because like I'm a professional Internet user,
and that's the first thing to learn is I don't
trust anybody. I wouldn't even want to hear it myself,
much less my kids, though I know.
Speaker 3 (29:38):
It like brings me back to like like when I
was like you know, eleven or twelve, Like the Internet
was like a new exciting thing, and I just would
like I remember going to like I like was like
really into like going to like like different video game
websites and stuff. And then there was these just these
chat rooms about video games you could go to and
(29:59):
it was literally just like a room with like everybody
their microphones are on and you just start talking and
it was like, when I think of that now, I'm like,
oh my god, that's frightening. Yeah, but uh yeah, I
mean the fact is is we can we could keep
those spaces safe with these sorts of tools, because then
(30:21):
then air Duell's out there at least get banned when
they're when they're being inappropriate or whatever.
Speaker 4 (30:28):
Oh god, Well, one thing that you put in the
show notes today that I would really like to hear
more about is a new speech to text engine or
engines cheetah and leopards. So maybe you could tell us
a little bit more about those.
Speaker 3 (30:41):
Yeah, so I think, yeah, last time we spoke, we
actually didn't have a publicly available speech to text engine,
and we were using our speech to intent engine, which
was called Rhino, which was basically, like you, basically, yeah,
the founder of the company is pretty obsessed with animals,
(31:02):
So Rhino basically you teach it a small grammar and
then it would understand that grammar, which is great for
stuff like you know, controlling a coffee maker or like
you know, there's only so many functions that needs to understand,
but we decided to kind of go that extra mile
and bring speech to text to devices. And traditionally language
(31:25):
models are in the gigabyte realm of size, and we
actually got ours down to twenty megabytes for language and
that's sort of the big win for this is like
we can run on anything that can take twenty megabytes
of memory or of storage. And so Leopard and Cheetah
(31:46):
are actually two different sides of the same coin. So
Leopard is a speech to text engine that takes in
a set amount of audio, so like an audio file
or something, and gives you a transcript of that, and
that's a lot that's an easier problem because you can
basically say, okay, this is all the audio I'm going
(32:06):
to get, so I'm to look forward, I'm going to
look back, I'm going to make inferences based on the
future in the past and give you a response. But
then Cheetah, of course, because it's the fast one, it
goes it's real time, so it has zero look ahead,
which means it will take in every frame of audio
that you give it and it will return what it
thinks is being said, so they're both speech to text engines,
(32:28):
but they just work at different use cases. So I
mean audio files. The accuracy is much better, but of
course you sacrifice the sort of real time effect.
Speaker 2 (32:38):
Yeah, so twenty megs is impressive, but is that still
like small enough for a browser to use, Like does
a user have to download that to use it in
their WebP.
Speaker 3 (32:47):
So that was a challenge we recently. So we recently
did the webstcs for Cheatah and Leopard, and we actually
had to kind of redesign our whole system of delivering
language to the the browser to handle this. So, yes,
twenty megabytes is a lot, but we actually separate the
language model from the package, so basically we let the
(33:10):
developer decide how that's delivered to the user, but we
also made it part of our system that it could
be either a basicxty four representation that you can bake
into your website if you just want it to always
be there, or if you want to be kind of
smarter about it, what you can do is put it
in your public folder and have it downloaded to the
user's browser on first load, and then cashed in local
(33:33):
storage for the rest of the time, so that the
next So the very first load, yeah, it'll be a
twenty megabyte load, but the second load will be instant
because they already have the language model.
Speaker 2 (33:44):
It's a pretty neat system because I think like it's
it's the nature of the beast, because I mean it's
it's in a way, it's kind of more of like
a native app feature, and native apps are downloading like
gigs at times of stuff, and so it's like a
feature that helps the web sort of compete with that.
So I think it makes sense, and I think, like
honestly that I think that's the best that this solution
(34:06):
is kind of clever because that's kind of all you
can do because you can't you can't magically get it
to the user ahead of time, like through an app
store or something.
Speaker 3 (34:15):
So and a developer, if they're if they want to
be clever about it, they can they can stream it
from their public folder asynchronously on the first load, so
that it's just like by the time the user wants
to activate the voice feature, it's already downloaded. You know.
It's it's just the sort of thing you need to
you need to handle these sorts of ways. Because you know,
(34:36):
we were working with a company recently that they do
this all the time in their mobile apps. They'll their
mobile app actually downloads like stuff all the time to
keep their their app working, and it does it all asynchronously,
like when you open up the app and you know,
the user's none the wiser, but behind the scenes there's
all this stuff. So when when you look up, why
(34:57):
is this app using three point six gigabytes? When I
downloaded it, it was only five hundred miniwtes, that's because they
only delivered like the core code and the rest of
it was downloaded later on. Yeah, so's it's it's just
how it's just how you do stuff now is just
keep keep the package sizes small, but then just deliver
(35:17):
the features kind of as they're being used.
Speaker 2 (35:19):
Yeah, I know, iOS and Andrew even have like APIs
built in to help you do that sort of thing
because it's it's such a common model.
Speaker 3 (35:27):
Yeah, I think all the big companies want that. You know,
if you're Spotify, you just you got to have the features.
You don't want people to see three point six gayabytes
when they go to download your app. There's like a
sticker shock thing that happens. So it's kind of a
funny thing because it ends up being that it's it's
like when you book like an Airbnb and there's all
(35:50):
these extra expenses that like get reported later, or like
a or a flight where you get like the info later.
It's sort of like that. It's like reduce the sticker
shock and then we will show you the expenses after.
Speaker 2 (36:03):
So you also have an article in here about writing
a podcast at transcription server to struggling to pronounce for
some reason, which is a fascinating idea that I think,
Like I know when we were talking before the show too,
we've done transcriptions and videos. I'm sure there's other people
that are call centers is another example, right, the things
that you want to transcribe. So does that use Peeter
(36:26):
Leopard or how does that work?
Speaker 3 (36:29):
Yeah, so it used this Leopard because we actually have
the ability to get a whole file, like an hour
long podcast and transcribe it from start to finish. And Yeah,
the reason I kind of came up with that as
an idea to sort of demo or technology is like,
I know, I've listened to podcasts for years, and like
it's so often on a long running show. I'm sure
(36:50):
on this show you get the Hey have we talked
about that? Did we talk about this? I feel like
we've talked about this, and having show notes to go
back too is probably a really helpful thing. Or like
I was thinking of doing a next phase of the
article where I actually make a podcast like searchable. So
I made it transcribable and basically stored the like text representation.
(37:14):
But once you have the text representation, you can make
it searchable and then you can start being like, oh,
when did I say this? And then it will just
pop up the episode you set it in. So it
was just kind of kind of an idea I came
up with because I see a lot of people using
Leopard on a server to basically hook into an event
that's happening somewhere, whether it be on ourrs feed, RSS
(37:37):
feed that's like updated or you know, yeah, like a
new audio file or video is uploaded, and it hooks
into that event, it runs it through Leopard and then
stores it in a database. I thought that was like
kind of a universal use case, Like it's just so
it seems like a fundamental part of the web to
like have something like that in a server.
Speaker 4 (37:57):
Yeah, I mean, it's it be so useful and it
would it would help I think everybody from people who
just want to reread part of a podcast if they're
looking for something specific instead of having to just kind
of hop through trying to figure out where it was
that that useful bit of information was well.
Speaker 3 (38:16):
And and you can think too, like you can deliver
these like let's say you attached it to your podcast.
You can like deliver these transcripts along with your podcast
because if you have the server hook in, transcribe it,
and then deliver the transcript along with the podcast, suddenly
you've got a follow along with the transcript podcast. So
(38:36):
these sorts of things are useful for like auto captioning
like videos or audio as well for like accessibility.
Speaker 2 (38:43):
It's accessible. It's also like marketing. People like it for
SEO purposes too, because you know audio, Google can index audio,
but if you have a transcription, it absolutely can even better.
Speaker 3 (38:54):
Yeah, one hundred percent correct. Yeah, search engines aren't very
good at indixing audio, so you just have to plaster
the text somewhere.
Speaker 2 (39:03):
Can you recognize different speakers because that's the other thing
about a transcript, right is knowing who's talking you do
you have the ability internally even if you don't know
names obviously, but can you say, like this is voice one,
voice two.
Speaker 3 (39:15):
Yeah. So actually we're working right now on a I'll
give you guys the scoop right now on a speaker
identification system that will basically be able to tell people
apart because yeah, when when you think of something like
doing like like a Zoom meeting, if you want like
to have meeting notes, Yeah, it would be really useful
(39:36):
to have like this came from this person, This came
from this person, This came from this person, and you
can use different I mean, Zoom obviously has the ability
to know where the audio is coming from, so it
can kind of just label it. But if you have
an anonymous audio stream with a bunch of different voices,
that's that's challenging because you don't know. You just have
to base your assumptions on the character of the voice.
(39:58):
Who's who's different. That's actually a problem we're working out
right now. And I mean that's that's useful for not
only speaker labeling, but also a speaker verification. So like
if we want to voice activate something that only responds
to your voice, that's also like another use case for it.
Speaker 4 (40:14):
That would be cool.
Speaker 2 (40:15):
Well, this has been a blast. Is there anything that
you wanted to discuss today that we have not gotten
to at all?
Speaker 3 (40:24):
No, I think I think we covered a lot here.
Speaker 2 (40:26):
Yeah, yeah, excellent. So that why don't we move into
our picks, and Paige, do you want to kick us off?
Speaker 4 (40:32):
Sure? So my pick is going to continue the trend
that I started last week, which was Star Trek. As
many of you who have been listening for a while,
I've been on a Star Trek journey through Next Generation
and forward. So most recently I've begun watching Star Trek
Lower Decks, which is their animated series, and there's only
(40:54):
I think there's maybe two, maybe three seasons of it,
but it is. It is in the stuff of Rick
and Morty, and it is the funniest Star Trek that
I've ever seen, to the point where I'm actually laughing,
which is unusual for anything animated, but it is really good.
There's a lot of references to other Star Trek franchises,
(41:17):
so if you are familiar with Next Generation or Voyager
or Enterprise, they throw in all sorts of little jokes
that are related to those characters. So I would definitely
recommend it. It's it's as family friendly as the rest
of the Star Trek a franchise is, and it's also
got a much bigger dose of humor than most of
(41:38):
them do. So if you're looking for something that's quick
twenty twenty five minute episodes, I would definitely say it's
a good one, excellent.
Speaker 2 (41:46):
I still have not gotten into the Star Trek world,
so it's at some point. I've had it recommended several times,
but I feel like it's such an like you can't
like casually wade into it, right, like you kind of
have to. Yeah, so if I pick this week is
going to be The Great British Bakeoff, which I think
was the previous of Year's page. I started off watching
(42:07):
it because I just wanted to know what it was about, right,
just that sort of thing. And then next thing I
knew I had watched a few episodes and I didn't
even really understand why.
Speaker 3 (42:16):
It's so it's so comforting that show, Like there's just
something so positive and warm about it.
Speaker 2 (42:22):
It's it's strangely compelling, Like I can't even understand why
I ended up watching it, but it's quite good. I
think Netflix has like five seasons or so, I mean
I don't know how much I'm going to watch, but
I've it's a good thing to just have on you're
not sure what to do. It's just comforting, nice to
have on in the background. So I've been sucked into
that as well.
Speaker 3 (42:42):
That's funny. My wife and I like literally just started
watching that like a few weeks ago, and for the
same reason, like why are people talking about this so much?
And yeah, now we're like shouting out the screen that's
not a good bake. Look at the problem mature.
Speaker 2 (42:57):
Yeah, excellent, Ian. What picks do you have for us?
Speaker 3 (43:03):
Yeah? So I think last time I brought Mandy a
really gnarly horror film. So I figure back up and
maybe do something a little different this time and actually
go with something tech related. So we've been working with
Mixed Panel recently, which is an amazing service. It's really
helped us because we're trying to add some analytics to
(43:24):
our website and our console, but custom analytics that allow
us to like basically track like basically when somebody enters
the website, what they interact with and like how long,
and like develop our own metrics based on the code
we actually put in the website and mixed Panel is
amazing at this. What they basically do is they're just
(43:44):
all they do is they say, hey, we're just going
to take events and we're going to represent them in
a whole bunch of different ways. You can filter, you
can form funnels, you can show userflow, you can take
a user and actually watch where they go on the
website and stuff. It's super super helpful for us. We've
actually been we've had a total crush on them since
(44:06):
we started working with their their product because they just
not only is their their UI like so nice to
work with, but it's just made our life. We were
thinking of building this far because basically we know wanted
data analytics to our website, but Google analytics and that
the sort of general analytics were just not enough. We
needed like really specific ones, right, and we were going
(44:27):
to build it ourselves, and then we stumbled the prompt
upon Mixed Panel and it was like, oh my god,
this saved us, Like like they made what we could
have made. It would have taken us. We would have
had to start a new company to make what they made,
and it's just it's so it's so helpful, So definitely
for any developer out there, that wants to add like
(44:47):
customer analytics, mixed panels, really really helpful.
Speaker 2 (44:51):
Awesome, excellent, Well this has been amazing. My last question
for you. If people want to follow you, keep up
with you, what are the best we're the best places
to go to do that.
Speaker 3 (45:00):
Yeah, I mean, let's see, I don't I don't have like,
I don't have like professional socials out there, but I
do have. I am on medium as Ian Lavery, so
you can read any articles I put up there. You
can follow Peak a Voice AI on Twitter, and we
have a YouTube channel if you want to check out
my bands. Yeah, fellow kids in Sleep Circle check them out.
Speaker 2 (45:23):
Yeah, no, excellent, that's great. We'll get those in the
show notes. And yeah, thanks for joining us today. This
is a great chat.
Speaker 3 (45:30):
Yeah, thanks for having me. This is great.
Speaker 2 (45:32):
Cool, all right, everybody until next week.
Speaker 4 (45:35):
See you then,
Welcome to React Roundup, the podcast where we keep you
updated on all things React related. This show is brought
to you by Void and top End Devs. Unvoid provides
high quality design and software development services on a client
friendly business model. Unlike all other software agencies, Unvoid allows
(00:20):
clients to only pay after the work is delivered and approved.
Visit unvoid dot com to learn more and reach out.
If you know a company that needs more professionals to
help with design and software development, that's u n void
dot com and top end Davs helps you stay up
to date with cutting edge technologies like JavaScript, Ruby, Elixir,
(00:44):
and AI. Visit topandevs dot com to join their AIDV
boot camp, weekly community meetups and access expert tutorials. I'm
Lucas Paganini, founder of Onvoid and host of this podcast.
Thank you for tuning in. Let's jump into the episode.
Speaker 2 (01:07):
Hey everybody, and welcome to another episode of React Roundup.
I am your host today TJ Van Tol and with
me on the panel, I have Paige need you House.
Hey everyone, and our special guest today is actually a
React Round of returning Champion. We have even Lovery here. Ian.
Welcome back to the show.
Speaker 3 (01:24):
Hey, thanks for having me back.
Speaker 2 (01:25):
Yeah, so why don't you start, you know, for people,
I think it's show we're looking back? Is the show
is about a year ago. We'll have to look up
the episode number and toss it in the show notes.
But it's been a while, So why don't you tell
people know who you are, what you do in your
background while you're famous, all those sorts of things.
Speaker 3 (01:41):
Yeah. So I work for a speech recognition company called
pegle Boys, and we're a developer focused company that tries
to power developers all over on any platform to have
to bring voice to their platform. So we have a
whole variety of different propus that cover speech to text,
(02:02):
voice activation, wake word, all that, and we just want
everybody to have a voice on their platform. Besides that,
I'm a I do like interactive media hard and I
play bass in a couple of bands.
Speaker 4 (02:16):
That's awesome, not just one band but multiple.
Speaker 3 (02:20):
Yeah, I'm an over at cheap I.
Speaker 2 (02:22):
Guess well cool. So Peka Voice looks interesting. I remember
us talking about it last time, but maybe you can
get an overview of like how it works. Like if
I if I use Peka Voice, what am I? What
am I getting? Am I getting a service that I
can send like audio to you, and it comes back
with the words like what other features maybe you could
give us, like the rundown of everything. It does everything
(02:43):
you do.
Speaker 3 (02:44):
Yeah. So the big thing with us is and our
sort of thing that sets us apart from pretty much
every other voice service is that we're entirely on device
and so there is no there is no service. There's
no cloud API that you're calling to send your audio to,
which I mean, look look around. That's pretty much every
(03:05):
single voice thing is just an API. So we're one
of the only ones out there that is actually giving
you the ability to hold on to your audio data
and your user's audio data and process it on the
device and return Again. We have like a variety of products.
So we have like wakeword detection, where it's just like hey,
(03:25):
Siri and okay Google. It's just all it's doing is
sitting there processing frames of audio, waiting for you to
say the thing, and then when it wakes up, it
does the thing that you tell it to do. But
we also have voice activity detection and which just basically
peaks when it hears somebody talking. And obviously speech to text.
Everyone wants speech to text, so auto transcription of voice.
Speaker 2 (03:49):
Yeah, it's very cool. It's also like one of those
problems that I feel like is it's becoming more commonplace.
We have smart devices in our house. Our phones can
listen to wake words and that sort of thing. But
I still I'm still sort of fascinated by the underlying technology.
Maybe you could just start give us like the world's
simplest rundown of like how does how does it actually
work on the back end? Like do you just have
(04:11):
a whole bunch of like low level C code that
looks for patterns in audio data or like, I don't know,
we don't need it. Sounds like two hours, but I'm.
Speaker 3 (04:20):
Just no, it's a good question. So, I mean, basically,
it's deep learning, right, It's it's it's machine learning. So
we teach through machine learning. We teach a machine a
statistical model of what a word sounds like, or what
a series of sounds sounds like. So we basically take
audio in our actual When we're teaching our machine, all
(04:41):
we're doing is sending it frames of audio that are labeled,
and we get it to remember them and like form
a little statistical pattern, and then it for something like wakeword.
It's just like, hey, remember this pattern of three things.
Just remember that and say, hey, I think I saw it.
So it's a lot more complicated when you get into
speech to text because not only are you teaching it
(05:02):
every sound in the language, but you're also teaching it
every word in the language, because then you're dealing with
audio and writing, which are different things. I think people
think language is a combination of those things, but really
they're two entirely separate things. They're like that there's the
(05:23):
series of sounds you make with your mouth that other
people understand, and then there's the symbols you write them
down with and the grammar and punctuation and everything that
you put into the written form, and they're different, so
we actually have to treat them differently. But you'll see
a lot of the big cloud providers out there. The
reason they got it so right so fast is because
(05:46):
they had such large machines in the cloud in order
to do this, so sort of like it outpaced the
actual progressive voice recognition, and now everything's kind of caught
up and we can actually do it on Devine, which
is a big win because, to be honest, we were
like boiling the ocean for like a while doing speech
(06:07):
to text, and now we can do it on like
a micro controller.
Speaker 4 (06:10):
So if you're using something like Peaco Voice, is it
something that you as a user have to train the
models or the models already there. It's trained. It knows
you're speaking English or it knows you're speaking Spanish, and
it will just it should be smart enough to be
able to take that audio and translate it into the
(06:32):
correct written words.
Speaker 3 (06:34):
Right, So, like for speech to text, for instance, we
basically just have a general language model. You just give it.
We offer eight different languages, and you just give it
the language you want and we'll understand that language. But
we actually use this thing called transfer learning, and we
have a website Peako Voice Console where you can basically
(06:56):
we have sort of a general model, but then you
sort of do train it yourself. Because for something like wakeword,
we have a model that understands a bunch of sounds
in whatever language you give it, but then you want
it to represent a certain series of sounds like okay, Google,
So you literally type that in to our console and
hit train, and then it will pop out a model
(07:18):
that understands that. So that's that's sort of the when
we say you train it, it's not like, oh, you
have to go out and gather four thousand recordings of
this word and you know, submit it to something and
watch statistics go and decide. No. No, it's just like
we are. We did the hard work.
Speaker 2 (07:38):
I was gonna say, because by saying that, you're sort
of implying that you got went out and have four
thousand recordings of these different words, right or like.
Speaker 3 (07:47):
No, No. So the thing is, it's again we've we've
trained the general model, so it understands the sounds we
needed to understand. You just tell us which sounds you
want us you want to form your wake word, and
we pop out a model that's that that just waits
for those series of sounds.
Speaker 2 (08:02):
Interesting because I would have guessed that your building of
the model was to get a bunch of people to
say like it almost seems it kind of breaks my
mind a little bit as possible, right, that you can
sort of general.
Speaker 3 (08:14):
Us the old style the like. So I worked. I
worked at a speech recognition company right out of college.
And what we did we had one of the early
early wakeword engines, and what we would do is we'd
it was all b to be the company. We basically
enter a contract with the company that says, hey, we're
(08:36):
going to go out and gather four thousand recordings of
this wake word, and we're going to train it and
then deliver you the model. And it was very formal,
and that was basically state of the art at the time.
But we're actually a bit past that now because we're
able to use this concept of transfer learning to take
(08:58):
a general model and just kind of pointed in the
right direction. So we no longer need to do all
that all that pounding the pavement asking for people to
say a wake word, because that was a lot of
work and it took months, like every time somebody signed
a contract. And I know because I was running the
crowdsourcing technology for that company, So I would have to
post these jobs and these these people would record it
(09:22):
on their on their like mobile device, and I'd have
to go through all the recordings and like you know,
some people would just yeah. Some people would just you know,
speak their manifesto into the phone, and I'd be like, no, no, no, no.
Speaker 4 (09:37):
So one one thing that I'm curious about is I'm
assuming that when you would do these these wake word gatherings,
you would have to take into account accents, because I
know that that is something that every automated assistant struggles with.
This English accents, Scottish accents, Caribbean accents, all speaking English,
(09:57):
but all slightly differently. So is PEKO voice able to
account for that and be able to interpret, you know,
a deep Southern accent versus maybe a New York Boston accent.
Speaker 3 (10:11):
Yeah, So I mean that that's still a challenge for us.
But I think the reason we're a bit more resilient
to it is because we've trained this general model on
like g'z like ten hundred thousand hours of speech. It's
heard all the accents, not not all the accents, but
it's heard it's heard a lot of variation, so it
(10:34):
tends to be a bit more resilient. When I was
doing the old style where we would get people to record,
that was actually a lot less resilient to it because
we only had like, you know, three hundred participants recording
these wake words, and how much variety are you going
to get between three hundred people? Like? Not enough? But
(10:55):
when we train these general models, we have like tens
of thousands of different speakers, maybe more, so we tend
to be a lot more sensitive to the variations. But
but it is, it is definitely a challenge because even
us as humans, if you hear like a really thick
accent that you're not used to, it can be confusing,
like like we're we're not perfect either with it. So
(11:18):
it's it's it's a challenge.
Speaker 2 (11:20):
So I think you so you added multiple language depart
I believe that's new or at least newish from the
last time we talk. So does that that like more generalizability,
make that easier or I imagine there's still all sorts of
challenges that go into that.
Speaker 3 (11:38):
Yeah, So when you when you actually work with a
totally different language, that's basically starting over because accents is
one thing you've already taught it the series of sounds
in the language, and you're just looking for a combination
of those sounds and those symbols. But when you move
into a new language, there's a new set of symbols,
and there's a new set of sounds. You know, there's
(12:00):
everybody has an inventory. We call it a phonemic inventory,
and it's basically a series of sounds that you hear
in the language, and every language has a different phonemic inventory,
and we need to train the machine to understand only
that inventory of sounds and all the symbols that go
into that. So when we start a new language, we
(12:21):
have to do it completely from scratch. We have to
get new data in that language, We need to get
new text in that language, and we need to do
our best to even understand the language enough to work
with it because we need to listen to these recordings.
We need to normalize the text we get and make
(12:44):
sure it's not like full of symbols and stuff, but
understand it enough so that we actually don't confuse the
machine learning process, and that that could be a real challenge.
It's a lot of work.
Speaker 2 (12:55):
Actually, so it's fascinating. Does that mean like when you
kick off anywe language, I feel like you almost need
to have like a professional linguist on staff for almost
each of these languages, right, Like, or do you like
bring on somebody who's, like, you know, a world class
I don't know, a Spanish linguist to help, or like
(13:15):
like how much of it are you able, like as
a software developer to sort of test on your own
and how much do you have to rely on a
native speaker as the only person that can actually figure
some of these things out.
Speaker 3 (13:26):
Yeah, so we do have like basically our like machine
learning team. They do have to be part linguist, like
because if you've studied languages, you at least understand the components,
and basically every language is just a combination of the components.
So they have a lot of expertise in that field
to understand when they approach a new language how it works.
(13:50):
But then that's not enough. So usually what we'll do
is we'll get somebody, well, we will get a native speaker.
Usually will basically hire somebody on a contract to work
with us to help with the language, because you do
need that expertise. Like, the fact is, even somebody who's
(14:11):
like a language expert, if they sit down to an
entirely new language, they're not going to be able to
understand it enough to do the work that needs to
be done to actually get it to a production ready state.
So we often do need to get a native speaker
in there to provide their input and that will really
(14:31):
speed the process along. We tried to do it without
experts a couple times, and it's just like you just
don't get the performance and you spend a lot more time,
you waste a lot more time. I should say sure.
Speaker 4 (14:44):
I mean that makes a lot of sense. When you
think about getting expertise in anything else, it's a lot.
It will almost undoubtedly go much quicker if you have
somebody who is proficient in whatever it is that you're
trying to do.
Speaker 3 (14:58):
Yeah, well, they can recognize mistakes, grammar and stuff, the
stuff that's really hard to pick up as a non
native speaker.
Speaker 4 (15:04):
Yes, So what languages do you currently offer Peko voice for?
Speaker 3 (15:09):
So we have I believe last year we announced we
had Spanish, French, German, English, and then this year we
added four new languages. We added Japanese, Korean, Portuguese, and Italian.
Speaker 4 (15:26):
These are some tough ones.
Speaker 3 (15:28):
Yeah, well, especially like when you get into the written
forms of Korean and Japanese, they become very challenging. Like
you know, we in English we have twenty six characters.
Japanese has two alphabets of fifty six and then an
additional alphabet of tens of thousands. So yeah, yeah, so
(15:53):
that the text representation of that is really difficult. The
actual spoken version of Japanese is a lot easier than
English because Japanese has fifty six sounds, and they all
map to a combination of characters. English mapping a combination
of characters to the sound is incredibly difficult. Turns out
(16:16):
we made some mistakes early on and we didn't really
fix them.
Speaker 4 (16:20):
I mean, just thinking about the amount of spellings that
we have for the same sounding word based on context,
I can not even imagine how you would be able
to figure that out for a transcript.
Speaker 3 (16:32):
And it's all exceptions in English. It's like, oh, yeah,
it's this unless this or this unless this, and like
here's three different reasons why this rule is wrong.
Speaker 2 (16:41):
And yeah, you see this when you have like younger
kids that are starting to write, and you look at
their writing because they start they don't know the exceptions yet, right,
but they can speak it because they know. So you
get like, they it's words you don't even think about too,
because we internalize them so quickly. Because one of my
kids spelled because ron and then you're like, oh, bill,
(17:02):
because it's pretty easy. But then you think about it
for like half a second and you realize, like, actually,
the word because makes absolutely no sense, like right.
Speaker 3 (17:09):
Like if you try and explain it, you suddenly find
yourself going just is what it is. Yes, just memorize it, yep.
Speaker 4 (17:19):
I mean that's really fantastic that you have taken on
and it sounds like gotten through some very difficult dialects.
Are what are future future languages that you hope to
be able to process as well?
Speaker 3 (17:32):
So we're going, yeah, exactly, So we're going to try
next year. We're going to double our language count again,
I think, and we're going to do going to do Chinese, Vietnamese,
what else? Dutch? I believe, Russian, Polish, I think, yeah,
I can't remember all of them. But you basically need
(17:52):
to be like a fully inclusive speech recognition company, you
basically need like a bare minimum of like fifty languages.
So like we're going to get to like twenty of
the most popular and hold there for a while. Is
kind of our plan because that covers a lot of people,
Like that covers the majority of people, because because even
(18:15):
in the cases where the people might not speak the language,
they usually are like, oh but I speak this what
this more popular language? But to really get up there,
like I mean, you do need to get to like
fifty or something. And I mean Google has like one
hundred and fifty, so you know, it's it's kind of
a never ending thing for us.
Speaker 4 (18:37):
How about Hindi that's a big one.
Speaker 3 (18:39):
Oh yeah, that's actually one of the other ones we're
going to do next year.
Speaker 2 (18:43):
Yeah, so I guess I got to ask one last question.
Are there any languages like you've come to hate, like
because it was like very difficult or.
Speaker 3 (18:56):
It's funny how much you can hate your own language. No, actually,
like seriously, English is the only Like I look at
all other languages we've done, and I'm like, these are
so much easier, like English is. Actually, it's just it
came out of a mess of languages. It was a
lot of combinations that happened over time, and a lot
(19:18):
of them happened during like you know, a lot of
English developed during like illiteracy, and so there's like really
interesting examples you can find of like stuff where it's
just like, oh, yeah, this was just a mistake that happened,
you know, two hundred years ago that they kept in
or actually, I have a fun fact the word dumb.
So you look at that, you're like, why does it
(19:39):
have a be at the end that apparently was there
was a time where the like ruling class of England
was trying to make it harder to write English so
that the peasantry could like pick it up. And they
literally just added some letters to the language here or there,
and we're like, this is the proper way to write it,
(19:59):
and then just to confuse people. And we literally still
have that to this day. So English is so weird.
Speaker 4 (20:06):
So that's why knife has a K in front of it.
Speaker 3 (20:09):
Yeah, yeah, like like stuff like that. I think they
were just messing with us and now we're just like
we have to live with that.
Speaker 2 (20:17):
So I want to pivot a little bit and talk
about the actual web development, like the side where you
might actually use a service like this, because I remember
last time we chat it a little bit too about
common use cases, right, So maybe we could just start
with a review, like how we have a lot of
web developers listen to this show. What do you think, like,
I guess, A, what would using something like this look like?
(20:39):
Like how do you actually get it in an app?
And B I guess like, what are some common use
cases that you see for use on the web as well?
Speaker 3 (20:46):
Right? So one of the big things is obviously on
the web, people are a lot more comfortable calling like
an API and that is what they've come to expect
for speech trek cognition and stuff. But we're actually bringing
the We're actually kind of bringing back the power of
the browser itself. So the I mean the browser is
a virtual environment that can run whatever you want. And
(21:09):
we actually can run entirely in the browser on the
client side. And that's that's big because I mean, in
the these days, we're getting a lot of progressive web apps,
and the sort of web app is a big thing,
especially with like SaaS companies and stuff. So if you're
running like a SaaS company and you you want to
(21:30):
integrate like voice into your console or something, having it
on the client side is is I mean, it lowers
the latency, It gives you a lot more direct control
of what happens when you get boys, and it means
you can be robust to connection issues, which like that
that you know, that's a huge thing. Not everyone has
(21:53):
amazing Internet and you don't want to have to be
I can calls out to an API and just hoping
it comes back for you or feature to work. This
will just work. And also on top of all that,
it's it's less expensive because we're not calling an API,
We're not depending on cloud infrastructure. So you're actually if
(22:13):
you're a developer and you integrate Peka Voice into your
web app, your client is going to be using their
machine to do the processing. So I think it's just
a win win situation for that. Yeah.
Speaker 2 (22:26):
I feel it's especially important considering it's audio too, So
like bandwidth is like you you're not just shipping off
like a couple of things in a query string to
some service. You're like uploading.
Speaker 3 (22:37):
Audio h mega adio.
Speaker 2 (22:40):
Yeah, so the bandwidth consideration is amplified significantly.
Speaker 3 (22:46):
Yeah. No, and it makes everything that that actually allows
for like something like real time audio. Real time audio
is very challenging to do for an API because you
basically need to stream it to the service and have
responses being streamed back. That's that's really expensive. That's like
a constant bandwidth issue. But when you're doing real time
(23:08):
audio and it's all running in your browser on the
client side, it's it's snappy and you can do things
that require timing and.
Speaker 2 (23:16):
Yeah, very cool. And I know I think I remember
from last time too that because one of the ways
you keep it snappy is it's not JavaScript code running
in the browser, right, it's your I don't remember your
exact tech stack, but I know you have some sort
of fancy way of doing that. Maybe you could walk
people through some of the magic and challenges of how
that works.
Speaker 3 (23:35):
Yeah, so our core code is in like C because
we we were trying to keep it as efficient and
snappy as possible. Now C code, and when you think
of C code next to React, you're like, how does
this even work? Like can these two ever talk? But
it turns out they can with WAM, And what we
do is we compile basically all our core code into
(23:58):
a WAHSM binary and then we ship that with our
like MPM package. So when you MPM install Peka Voice,
part of the part of what's going to be shipped
with your website is OURASM blob. And basically wasm's really
cool because it basically just wraps your native code in
(24:21):
JavaScript and then allows you to basically attached to it
like any sort of dynamic library, say, here's the functions
I want to call, here's the data I'm going to
put into it, and then you just call it like
you would any any other library. It's a little trickier
(24:41):
to work with because you're dealing with I mean JavaScript
obviously one of the things we're pretty aware of, and
I'm sure the listeners of your show are aware of.
Is JavaScript is like eh, types whatever, even typescript is
like is like yeah, types, but like you know, a
number is a number, right? Well, see is like what
how many bits is your number? Like he needs to know?
(25:06):
So you start to need to think about that. When
you work with WASM, you start to need to think
about okay, is this a thirty two bit and going
in here? And you need to start to think of okay,
you need to start to think of memory, like okay,
I need to have a pointer. I need to pass
in a pointer here to get something back and then
convert that pointer to like a JavaScript object of some sort.
(25:30):
So it's challenging to work with, but once you get
it working, it's extremely powerful because then we can ship
something that's incredibly complex piece of code and just put
basically a slim interface of JavaScript around it, and then
any jabscript developer can just call is just talking to
(25:51):
it like it's JavaScript. They don't need to worry about
the WAHSM that was our problem. Yeah, So it's challenging
to work with, but I do if anybody's thinking of
has a challenging problem that requires the efficiency of c
Don't be afraid of it. It's not that it's not
that hard, and it is pretty amazing when you start
(26:12):
working with it.
Speaker 4 (26:12):
Actually, Okay, so it works or there there is an
NPM package if you want to use JavaScript with it.
But what if you are a Python developer or maybe
you're working with a micro controler like our do we
know is there are there options for other other languages
like that?
Speaker 3 (26:32):
Yeah, So, I mean we support since we're a developer
focused company, we're pretty obsessed with our SDKs. So I
think for our two most popular products, I think we
have like twenty SDKs for each one, and it covers
all the favorites. And we even have you know, we
(26:53):
have three No, we have four different web SDKs. We
have Vanilla JavaScript, but we also have React, Angular and view.
So it allows we basically wanted to be like, use
it in your favorite environment, like, yeah, use it like
you use anything else in your in your stack, Like
(27:14):
we don't want to just disturb that basically.
Speaker 4 (27:17):
Right, that's awesome. So what are some of the use
cases that you've seen people employing it for recently?
Speaker 3 (27:25):
So we've seen, so we've actually come to some interesting
ones lately. So auto content moderation is a big one
right now. So let's say you're Minecraft or something, or
you're I guess, let's let's go Fortnite, and you have
open audio streams hundreds of thousands of players, and you're
(27:48):
trying to moderate all that. That's that's very difficult. And
it turns out a lot of big companies out there
are using auto moderation, which basically takes that audio and
is basically looking for key phrases let's call them, and
(28:10):
it's just looking to flag them and then and then
they'll usually have you know, a person go in and
inspect the actual content of it and decide whether it
was you know, a mistake or whether it is actually
like a bannable offense. Yeah, so that that's actually a
new exciting one. Also, like call centers, it turns out,
(28:31):
again we've got open phone lines, like like a whole
building full of them, and we're trying to understand, you know,
what's being said on all these different calls, and you
can't have people listening to all that audio. So a
lot of big call center companies need some sort of
automated system to take in all the audio from all
their phones and do something with it. So we're we're
(28:54):
encountering more use cases like that lately.
Speaker 2 (28:57):
Actually, those are both really fascinating. It's funny. The content
moderation one really resonated with me because I play I
don't know. My kids are eleven, so they're right at
that impressionable age, but they're also right in the age
where they want to play like games that are the
sort where they have open audio. So there's a game
we play that's like five y five, so five people
(29:18):
on each team, and it has it has a way
for you to do audio communication. And the very first
thing I did was make sure to shut that off,
like disable it, because like I'm a professional Internet user,
and that's the first thing to learn is I don't
trust anybody. I wouldn't even want to hear it myself,
much less my kids, though I know.
Speaker 3 (29:38):
It like brings me back to like like when I
was like you know, eleven or twelve, Like the Internet
was like a new exciting thing, and I just would
like I remember going to like I like was like
really into like going to like like different video game
websites and stuff. And then there was these just these
chat rooms about video games you could go to and
(29:59):
it was literally just like a room with like everybody
their microphones are on and you just start talking and
it was like, when I think of that now, I'm like,
oh my god, that's frightening. Yeah, but uh yeah, I
mean the fact is is we can we could keep
those spaces safe with these sorts of tools, because then
(30:21):
then air Duell's out there at least get banned when
they're when they're being inappropriate or whatever.
Speaker 4 (30:28):
Oh god, Well, one thing that you put in the
show notes today that I would really like to hear
more about is a new speech to text engine or
engines cheetah and leopards. So maybe you could tell us
a little bit more about those.
Speaker 3 (30:41):
Yeah, so I think, yeah, last time we spoke, we
actually didn't have a publicly available speech to text engine,
and we were using our speech to intent engine, which
was called Rhino, which was basically, like you, basically, yeah,
the founder of the company is pretty obsessed with animals,
(31:02):
So Rhino basically you teach it a small grammar and
then it would understand that grammar, which is great for
stuff like you know, controlling a coffee maker or like
you know, there's only so many functions that needs to understand,
but we decided to kind of go that extra mile
and bring speech to text to devices. And traditionally language
(31:25):
models are in the gigabyte realm of size, and we
actually got ours down to twenty megabytes for language and
that's sort of the big win for this is like
we can run on anything that can take twenty megabytes
of memory or of storage. And so Leopard and Cheetah
(31:46):
are actually two different sides of the same coin. So
Leopard is a speech to text engine that takes in
a set amount of audio, so like an audio file
or something, and gives you a transcript of that, and
that's a lot that's an easier problem because you can
basically say, okay, this is all the audio I'm going
(32:06):
to get, so I'm to look forward, I'm going to
look back, I'm going to make inferences based on the
future in the past and give you a response. But
then Cheetah, of course, because it's the fast one, it
goes it's real time, so it has zero look ahead,
which means it will take in every frame of audio
that you give it and it will return what it
thinks is being said, so they're both speech to text engines,
(32:28):
but they just work at different use cases. So I
mean audio files. The accuracy is much better, but of
course you sacrifice the sort of real time effect.
Speaker 2 (32:38):
Yeah, so twenty megs is impressive, but is that still
like small enough for a browser to use, Like does
a user have to download that to use it in
their WebP.
Speaker 3 (32:47):
So that was a challenge we recently. So we recently
did the webstcs for Cheatah and Leopard, and we actually
had to kind of redesign our whole system of delivering
language to the the browser to handle this. So, yes,
twenty megabytes is a lot, but we actually separate the
language model from the package, so basically we let the
(33:10):
developer decide how that's delivered to the user, but we
also made it part of our system that it could
be either a basicxty four representation that you can bake
into your website if you just want it to always
be there, or if you want to be kind of
smarter about it, what you can do is put it
in your public folder and have it downloaded to the
user's browser on first load, and then cashed in local
(33:33):
storage for the rest of the time, so that the
next So the very first load, yeah, it'll be a
twenty megabyte load, but the second load will be instant
because they already have the language model.
Speaker 2 (33:44):
It's a pretty neat system because I think like it's
it's the nature of the beast, because I mean it's
it's in a way, it's kind of more of like
a native app feature, and native apps are downloading like
gigs at times of stuff, and so it's like a
feature that helps the web sort of compete with that.
So I think it makes sense, and I think, like
honestly that I think that's the best that this solution
(34:06):
is kind of clever because that's kind of all you
can do because you can't you can't magically get it
to the user ahead of time, like through an app
store or something.
Speaker 3 (34:15):
So and a developer, if they're if they want to
be clever about it, they can they can stream it
from their public folder asynchronously on the first load, so
that it's just like by the time the user wants
to activate the voice feature, it's already downloaded. You know.
It's it's just the sort of thing you need to
you need to handle these sorts of ways. Because you know,
(34:36):
we were working with a company recently that they do
this all the time in their mobile apps. They'll their
mobile app actually downloads like stuff all the time to
keep their their app working, and it does it all asynchronously,
like when you open up the app and you know,
the user's none the wiser, but behind the scenes there's
all this stuff. So when when you look up, why
(34:57):
is this app using three point six gigabytes? When I
downloaded it, it was only five hundred miniwtes, that's because they
only delivered like the core code and the rest of
it was downloaded later on. Yeah, so's it's it's just
how it's just how you do stuff now is just
keep keep the package sizes small, but then just deliver
(35:17):
the features kind of as they're being used.
Speaker 2 (35:19):
Yeah, I know, iOS and Andrew even have like APIs
built in to help you do that sort of thing
because it's it's such a common model.
Speaker 3 (35:27):
Yeah, I think all the big companies want that. You know,
if you're Spotify, you just you got to have the features.
You don't want people to see three point six gayabytes
when they go to download your app. There's like a
sticker shock thing that happens. So it's kind of a
funny thing because it ends up being that it's it's
like when you book like an Airbnb and there's all
(35:50):
these extra expenses that like get reported later, or like
a or a flight where you get like the info later.
It's sort of like that. It's like reduce the sticker
shock and then we will show you the expenses after.
Speaker 2 (36:03):
So you also have an article in here about writing
a podcast at transcription server to struggling to pronounce for
some reason, which is a fascinating idea that I think,
Like I know when we were talking before the show too,
we've done transcriptions and videos. I'm sure there's other people
that are call centers is another example, right, the things
that you want to transcribe. So does that use Peeter
(36:26):
Leopard or how does that work?
Speaker 3 (36:29):
Yeah, so it used this Leopard because we actually have
the ability to get a whole file, like an hour
long podcast and transcribe it from start to finish. And Yeah,
the reason I kind of came up with that as
an idea to sort of demo or technology is like,
I know, I've listened to podcasts for years, and like
it's so often on a long running show. I'm sure
(36:50):
on this show you get the Hey have we talked
about that? Did we talk about this? I feel like
we've talked about this, and having show notes to go
back too is probably a really helpful thing. Or like
I was thinking of doing a next phase of the
article where I actually make a podcast like searchable. So
I made it transcribable and basically stored the like text representation.
(37:14):
But once you have the text representation, you can make
it searchable and then you can start being like, oh,
when did I say this? And then it will just
pop up the episode you set it in. So it
was just kind of kind of an idea I came
up with because I see a lot of people using
Leopard on a server to basically hook into an event
that's happening somewhere, whether it be on ourrs feed, RSS
(37:37):
feed that's like updated or you know, yeah, like a
new audio file or video is uploaded, and it hooks
into that event, it runs it through Leopard and then
stores it in a database. I thought that was like
kind of a universal use case, Like it's just so
it seems like a fundamental part of the web to
like have something like that in a server.
Speaker 4 (37:57):
Yeah, I mean, it's it be so useful and it
would it would help I think everybody from people who
just want to reread part of a podcast if they're
looking for something specific instead of having to just kind
of hop through trying to figure out where it was
that that useful bit of information was well.
Speaker 3 (38:16):
And and you can think too, like you can deliver
these like let's say you attached it to your podcast.
You can like deliver these transcripts along with your podcast
because if you have the server hook in, transcribe it,
and then deliver the transcript along with the podcast, suddenly
you've got a follow along with the transcript podcast. So
(38:36):
these sorts of things are useful for like auto captioning
like videos or audio as well for like accessibility.
Speaker 2 (38:43):
It's accessible. It's also like marketing. People like it for
SEO purposes too, because you know audio, Google can index audio,
but if you have a transcription, it absolutely can even better.
Speaker 3 (38:54):
Yeah, one hundred percent correct. Yeah, search engines aren't very
good at indixing audio, so you just have to plaster
the text somewhere.
Speaker 2 (39:03):
Can you recognize different speakers because that's the other thing
about a transcript, right is knowing who's talking you do
you have the ability internally even if you don't know
names obviously, but can you say, like this is voice one,
voice two.
Speaker 3 (39:15):
Yeah. So actually we're working right now on a I'll
give you guys the scoop right now on a speaker
identification system that will basically be able to tell people
apart because yeah, when when you think of something like
doing like like a Zoom meeting, if you want like
to have meeting notes, Yeah, it would be really useful
(39:36):
to have like this came from this person, This came
from this person, This came from this person, and you
can use different I mean, Zoom obviously has the ability
to know where the audio is coming from, so it
can kind of just label it. But if you have
an anonymous audio stream with a bunch of different voices,
that's that's challenging because you don't know. You just have
to base your assumptions on the character of the voice.
(39:58):
Who's who's different. That's actually a problem we're working out
right now. And I mean that's that's useful for not
only speaker labeling, but also a speaker verification. So like
if we want to voice activate something that only responds
to your voice, that's also like another use case for it.
Speaker 4 (40:14):
That would be cool.
Speaker 2 (40:15):
Well, this has been a blast. Is there anything that
you wanted to discuss today that we have not gotten
to at all?
Speaker 3 (40:24):
No, I think I think we covered a lot here.
Speaker 2 (40:26):
Yeah, yeah, excellent. So that why don't we move into
our picks, and Paige, do you want to kick us off?
Speaker 4 (40:32):
Sure? So my pick is going to continue the trend
that I started last week, which was Star Trek. As
many of you who have been listening for a while,
I've been on a Star Trek journey through Next Generation
and forward. So most recently I've begun watching Star Trek
Lower Decks, which is their animated series, and there's only
(40:54):
I think there's maybe two, maybe three seasons of it,
but it is. It is in the stuff of Rick
and Morty, and it is the funniest Star Trek that
I've ever seen, to the point where I'm actually laughing,
which is unusual for anything animated, but it is really good.
There's a lot of references to other Star Trek franchises,
(41:17):
so if you are familiar with Next Generation or Voyager
or Enterprise, they throw in all sorts of little jokes
that are related to those characters. So I would definitely
recommend it. It's it's as family friendly as the rest
of the Star Trek a franchise is, and it's also
got a much bigger dose of humor than most of
(41:38):
them do. So if you're looking for something that's quick
twenty twenty five minute episodes, I would definitely say it's
a good one, excellent.
Speaker 2 (41:46):
I still have not gotten into the Star Trek world,
so it's at some point. I've had it recommended several times,
but I feel like it's such an like you can't
like casually wade into it, right, like you kind of
have to. Yeah, so if I pick this week is
going to be The Great British Bakeoff, which I think
was the previous of Year's page. I started off watching
(42:07):
it because I just wanted to know what it was about, right,
just that sort of thing. And then next thing I
knew I had watched a few episodes and I didn't
even really understand why.
Speaker 3 (42:16):
It's so it's so comforting that show, Like there's just
something so positive and warm about it.
Speaker 2 (42:22):
It's it's strangely compelling, Like I can't even understand why
I ended up watching it, but it's quite good. I
think Netflix has like five seasons or so, I mean
I don't know how much I'm going to watch, but
I've it's a good thing to just have on you're
not sure what to do. It's just comforting, nice to
have on in the background. So I've been sucked into
that as well.
Speaker 3 (42:42):
That's funny. My wife and I like literally just started
watching that like a few weeks ago, and for the
same reason, like why are people talking about this so much?
And yeah, now we're like shouting out the screen that's
not a good bake. Look at the problem mature.
Speaker 2 (42:57):
Yeah, excellent, Ian. What picks do you have for us?
Speaker 3 (43:03):
Yeah? So I think last time I brought Mandy a
really gnarly horror film. So I figure back up and
maybe do something a little different this time and actually
go with something tech related. So we've been working with
Mixed Panel recently, which is an amazing service. It's really
helped us because we're trying to add some analytics to
(43:24):
our website and our console, but custom analytics that allow
us to like basically track like basically when somebody enters
the website, what they interact with and like how long,
and like develop our own metrics based on the code
we actually put in the website and mixed Panel is
amazing at this. What they basically do is they're just
(43:44):
all they do is they say, hey, we're just going
to take events and we're going to represent them in
a whole bunch of different ways. You can filter, you
can form funnels, you can show userflow, you can take
a user and actually watch where they go on the
website and stuff. It's super super helpful for us. We've
actually been we've had a total crush on them since
(44:06):
we started working with their their product because they just
not only is their their UI like so nice to
work with, but it's just made our life. We were
thinking of building this far because basically we know wanted
data analytics to our website, but Google analytics and that
the sort of general analytics were just not enough. We
needed like really specific ones, right, and we were going
(44:27):
to build it ourselves, and then we stumbled the prompt
upon Mixed Panel and it was like, oh my god,
this saved us, Like like they made what we could
have made. It would have taken us. We would have
had to start a new company to make what they made,
and it's just it's so it's so helpful, So definitely
for any developer out there, that wants to add like
(44:47):
customer analytics, mixed panels, really really helpful.
Speaker 2 (44:51):
Awesome, excellent, Well this has been amazing. My last question
for you. If people want to follow you, keep up
with you, what are the best we're the best places
to go to do that.
Speaker 3 (45:00):
Yeah, I mean, let's see, I don't I don't have like,
I don't have like professional socials out there, but I
do have. I am on medium as Ian Lavery, so
you can read any articles I put up there. You
can follow Peak a Voice AI on Twitter, and we
have a YouTube channel if you want to check out
my bands. Yeah, fellow kids in Sleep Circle check them out.
Speaker 2 (45:23):
Yeah, no, excellent, that's great. We'll get those in the
show notes. And yeah, thanks for joining us today. This
is a great chat.
Speaker 3 (45:30):
Yeah, thanks for having me. This is great.
Speaker 2 (45:32):
Cool, all right, everybody until next week.
Speaker 4 (45:35):
See you then,
Popular Podcasts
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!