Free audio post production

by Alfonic.com I.

Know from the last time I just thanks the phoning,

I'll try.

Well, this maybe as good like probably represents to the sort of audio.

Yeah,

background noise.

Yeah, so this is our favorite question, which is like when you're building I mean, the building with voice AI,

if you could wave a magic wand at any of the problems that you've been dealing with, what would you wave it at?

That's interesting.

The...

Well, there's two.

So, I mean, the one we've been talking about, like the audio pinning out or whatever it is, because there's just such diverse audio sources that it's...

We don't know what're going to get, and you're going enough fusarely me can sell themload the vision and then you get to the store and it's just screeches and all them on you goes clipp to like,, who are going do this?

and so it was definitely that problem.

Um, the other one that is like very radio specific is Don't get your hands on Anderson Radio and this like it's a barrier for every sales they send us your radios and you're gonna get two of them, and' better have chargers, and like I don't know.

Okay, well, now we can't do that or we can't spare them.

So we're sourcing ourselves and like there's a bit of a hardware challenge there.

That'll go away eventually when we've done enough of these.

But the stage rad, like

jumping for joy when there's like, they got the same radios as we've already compatible whiles in there.

Yeah.

Or it's a cheap one that we could just buy and it's like, we're just heading to just go get one for $100 and install it for them on the ship.

Yeah.

Those two, the, the band of my existence right now.

So the voice, the, I'm just like.

The, the carrying it accuracy.

Carrying it accuracy.

Yeah.

Yeah.

Yeah.

Actually having the hardware to plug in.

So it would be, yeah.

And get it.

And that's so that you can actually develop.

You can only deliver it smaller.

Like we can do all the development.

We got a bit of a rude awakening because we did a bunch of the development and just been like, just use the speaker on your laptop and, yeah, I'll talk to it.

It basically gets what I say almost every time correctly.

And then that's not the experience we need.

Right.

Plug it through the hardware kit.

That's a walkie talkie.

How was it different?

Oh, it's just like you're going to get half as many successful returns from the speech to text.

It just comes back with nothing.

Like, couldn't interpret it or wildly incorrect.

Because of the- It's just such a bad signal.

Bad signal.

I'm not- was it the-

The Owookies are fine because they they don't have a lot of bandwidth.

So they got to compress the normal range of human voices.

They're going to throw out bits that they don't think you're going to need.

They know you're going to play it on another motorized Owookie.

Talking to the tone.

They got a little speaker this big.

So well, let's don't send any bass with that.

They're going to like,

yeah, make the audio worse so that they can fit it across the air.

And then like we've just gotten that signal to work with on the other end.

And that's one of the reasons we're looking at using something like a software defined radio to like try and exclude any artifact that's coming from the walkie-talkie actually playing the audio for us.

That's making assumptions about what kind of speaker it's got in it.

It's filtering out some of the audio anyway.

Cause it was always submodular with some missing things that bowled to the air that we want to prove whether like it's the walkie talkie itself that we're listening to the problem versus like what's coming over here.

That's a bloody project that we're looking to kick off.

And so it's like what you build, you need to actually test it on the devices.

So yeah, know that it's actually good to use.

Yeah.

None of our, a few of our customers are local, but we're sending these kits across the country.

If we don't have installers there, and say, oh, it's kind of just ship it to the manager of the store and they're going to plug it in the ethernet and power.

Hopefully it fires up.

They got to turn on the radios at volume directly, put it on the right channel.

There's a lot of them dumb little human problems that could go wrong.

And so what the magic wand would be to.

Infinite supply of devices.

Infinite supply of devices.

Yeah.

Yeah, that would solve that.

So if we could, if this wave of the magic wand worked, how would it change your life?

Now, I spend a lot of our time on the delivering.

Side, making sure that, okay, this kit that we build is actually going to work for this customer as opposed to like writing code to stupid new use cases in.

So that problem would just go away.

And would be dramatically less.

Yeah.

Good.

That's, this could also be solved with like a bigger production and testing team.

Yeah.

That would solve that too.

If you had a bigger, but yeah.

Just other people that aren't the dev team.

How to put on that stuff.

We just had that, baby.

Yeah.

Is there any, if there was like a third, is there a third thing that you would wave a magic wand at?

A third.

Sorry, a third thing.

I say, oh.

Short of like customer side problems or like IT and can we get approval to do this?

There's so many things that hang out, but the technology is fine.

It's gonna work while I'm getting blocked because,

yeah, the IT department doesn't want about network access or people are paranoid that the audio that we're capturing is somehow sensitive and that's the company's secrets, even though it's from public airways and someone could sit in the parking lot with a radio.

Yeah.

It's kind of like the human understanding of probably top of it.

Okay.

This is like human, it could go either way.

Yeah.

I think there's a fourth thing would be,

it doesn't need a magic wand, but just better, more time to devote to how we orchestrate the various modules.

Like, we're so independent right now.

Like, if you're doing an inventory lookup, you need to tell us it's inventory and then we can go do that really well.

And if it's a generic LLM, we can do that.

We can hand that off really well.

But then if you start mixing these things together, breaks up the LLM conversation.

Yeah.

You don't want to have to like ask the agent, Hey, Toby, every single time I talk to it, you want to be able to like ask it once and then if I keep coming back, you know, it's me and, that isn't handled very well because there's so much other traffic going on.

Oh, you can't really piece together the...

Somebody got a quick question in here that isn't for the agent, but then let's go back to answering the original one.

That was...

So it's like kind of keeping the consistency across.

Like, it's like you want to treat it as if it's like a conversation.

But this thing has like many conversations going on.

Kind of like if Granola was like, if we wanted to pick up the neighboring conversation.

Yeah.

At the same time, I keep it separate.

If it could do that.

That was blowing my mind.

Yeah, that's kind of the same problem, so basically that's the same.

Yeah.

That's very interesting.

Do you remember my third question?

I have one more question that I always use, which is, just pull out.

The top of my head.

Do you have more questions, Aiden, while I search?

I frantically searched for my third question.

I have a complete aside to that point, which is, I was just thinking, I was just saying that what technology exists that can do something like that?

And I did hear something, I thought it was really cool the other day.

I don't know if this would actually solve that.

It's like how you...

How do you breathe and shut your sounds, baby, and then be able to direct the microphone digitally off in the back?

Like there are apparently these directional 360 microphones.

Where are your ears?

Have you seen those?

Yeah.

Fucking crazy.

So you,

you could wear them in your ears and then it reports how it's in your ear that it didn't hear it.

That's good.

And apparently it's great for like reconstructing like soundbox kind of, but also has some real utility if you're like, I want to record this and I don't know where I want to point the microphone, but later I will figure that out.

Nice.

So yeah, that's my point.

Complete the side there.

That was very helpful, just so I now remember that.

A friend of mine that we talked to about, he's like a digital schooling and digital signal processing.

Oh, really?

So we've been kind of tapping him for some insights into what we're doing and kind of educated me on like how some of these things work, but like your Sonos system, like when you plug it in, it's gonna do these sign sweeps, like these big woops sounds like a few times.

I have some speakers that do that.

Basically gonna make a picture they can hear your room and they-- Really?

It's not really-- It's not six feet.

Yeah, it's crazy.

So they can understand how far that wall is away and where you need to be to hear the sound and they get reconfigured.

So the speakers, Jack's staying at my place while he's here.

The speakers in there, the big-- oh.

Yeah, the corners just-- They do that.

Really?

Yeah, if you turn one, they-- Then.

They bounce the sound around.

Yeah, it's cool.

It's like, amazing.

Sadly, it's never gonna work for a walkie-talkie.

But it kind of got up to the thoughts thinking around, well, what's a-- It's a way to like deploy a system and be able to test, is it working?

Or like what radio should you replace?

This is never gonna work.

And some tests that we could run.

I go grab every radio and play audio through it.

Okay, we can pick that out and identify the good ones and the bad ones and maybe adjust.

Yeah, it's so cool.

It's like bats.

Right?

Walk away, son on.

I got my last question.

Well, maybe last.

So if you were, so if you waved your magic wand and you solved, you made it, you know, no devices and you were able to like understand the signal well,

is there anything that makes that more valuable now than it was like a year or two ago?

I mean, the end.

Everything we're doing is to unlock data that is just out there and disappears.

Every one of these publications was the who we're trying to sell to.

And there's a ton of value in that data that was just vanished into the

ether.

So that's always been there.

It's just you've never been able to get at it.

So storing all that data and then starting to extract from it,

I think we're very much in the retail space and the things

thousands of little, little bits of data that are valuable at the kind of corporate level.

Yeah.

And then all kinds of tactical things at a store level, but let me help me do my job here.

But at the corporate level, it's like, what are the patterns and where's the waste?

And, you know, where do we get, yeah, what type of store performs better?

Or, the world starts doing certain things on the radio, like,

it was so much there.

So the value's always been there.

There's just no way to get at it.

In that end.

We haven't explored much beyond retail.

The number of places that radios used,

concerts, events, things like that, that have-.

Concerts is like taking the difficulty to 11.

Yeah, let's pick that noise reduction problem.

Oh, yeah.

Screaming.

Sure.

The humans understand what's going on.

Well, that's what I was saying.

Yes.

At some point, you know, we could.

Build stuff that's better than human senses, right?

So like, but yeah, they I mean, they're dealing with all kinds of problems.

It might be food and beverage or it might be security.

And like, yeah, they had a baseball game or something.

I was thinking that, like, yeah, music concerts and basic.

Yeah.

Yeah.

It's huge.

And they're, I'm sure there are systems that record all that.

Like, obviously they've got them for military and aviation applications.

They just record all the traffic, but not really at a civilian level.

And I don't know, I can see the point where But once this exists and it's out there, you're gonna get sued if you don't have it.

As far as like, you're not doing your due diligence to make sure your staff aren't doing things they shouldn't on the radio.

Sure, sure, sure, sure.

Yeah, if you're not monitoring your traffic, then you're truly in trouble.

You guys can become paladin for radio, basically.

How have you been learning about voice AI stuff?

Just kind of on the fly.

Kind of hacking through it.

Um, there's no courses or anything that I don't think we can take on.

Just APIs and blockchains and just, try it out.

Yeah.

I mean, we all lose the, the front ends.

Yeah.

I have a day in the agency.

Just throw himself like a copilot or chat GPT.

So kind of know what to ask and do.

The prompt engineering side of it, at least.

Well, getting there, but they put the APIs to talk to these systems, aren't they?

Yeah.

Yeah.

For sure.

I think it's all giant companies that have thought through security and they've got that figured out.

Yeah, it's nothing too weird.

Yeah.

That's.

Just something you don't think about.

Cool.

Cool.

Incredibly helpful.

Yeah.

Yeah.

Thank you.

Thank you very much.

Still recording.

Let's just let them go on.

Free audio post production

by alphonik.com.