Sound (featuring audio wizard Mat Keselman)

0:07Speaker 1:How any extraordinaries. Today, we're pulling back the dKu curtin a bit. We're gonna see how the sausage is made, as it were. So Daniel and I write pretty darn detailed outlines before we record an episode, but often something happens that totally derails our outline conversation, and that's good. Sometimes the derailment happens because something I thought was kind of obvious turns out to not be so obvious, and Daniel will ask clarifying questions that forced me to come up with different, better explanations on the fly. Or Daniel will be working through his outline and I'll think of some tangentially related topic that I now really want to know about, and Daniel will hop online to do a quick bit of research, and then we'll record a completely unplanned segment. In each of these cases, our answers are way less polished than we'd like, Lots of false starts, lots of basically lots of stuff that the audience definitely doesn't want to have to sit through. And then Daniel and I have unexpected noises that intrude upon the audio.

1:10Speaker 1:In today's episode, I learned that my microphone actually does pick up the sounds my goats make when they're trying to get my attention, and sometimes Daniel apparently records from the inside of a grocery store. More on that later. Anyway, these are only a handful of the many challenges that we throw at Matt Kesselman, the audio wizard who makes Daniel and Kelly's Extraordinary Universe sound so good. So today we're bringing Matt on the show to talk about sound and how he goes about fixing hours so that your listening experience is so much better that it would be if you were a fly on the wall while Daniel and I record. Welcome to Daniel and Kelly's Edited Universe. Hi.

2:00Speaker 2:I'm Daniel. I'm a particle physicist and I like thinking about aliens, and I've often been told I have a face for radio.

2:13Speaker 1:Hi. I'm Kelly leader Smith. I study parasites in space. I often make jokes about having a face for radio. But I think actually Daniel and I are but quite attractive human beings. Oh yes, and Daniel has a great voice for radio, So let's focus on the positive.

2:29Speaker 2:As do you, Kelly. I think our voices work really well together.

2:32Speaker 1:I appreciate it. But I will note that when we transitioned from Jorge on the show to Kelly on the Show, we did get a fair number of folks writing in to complain about the sound of my voice. But here I am over one hundred episodes later. Kiss my butt, loosers. I'm not going anywhere hater's gonna hate.

2:53Speaker 2:But we also got a lot of comments from people who love your voice. And I think it's just a subjective thing, you know, And it's a big part of finding a podcast you'd like listening to. Is are the voices pleasant and soothing? And that's very personal?

3:06Speaker 1:You know, That's right. But the good news is you and I have Matt Kesselman to help us whenever our voices are sounding less than their best.

3:15Speaker 2:That's right. And I want my voice to be smooth, because a lot of folks write in and tell me they like to fall asleep to my voice, and hey, you know, I don't want to disturb anybody's slumber. No.

3:25Speaker 1:I wonder what my Calkuli laughed as to their dreams. I dream of witches. But today we decided that we were going to invite Matt Kesselman, who is just an absolutely amazing audio engineer who, thankfully for us, spends a lot of time working on our show. And part of why we decided to invite him is because anyone who submits a question that ends up on our Listener Questions episode gets a raw audio file of Daniel and I having our conversation, and we send it to them before it goes to Matt, our audio engineer, so that they have a chance to record themselves asking the question and record themselves responding to our response to their question, and then we can send all of that to Matt and he can edit it all at once. And people often respond to us and say, wow, your audio engineer has a lot.

4:10Speaker 2:Of work to do, but not just that. Also they say that it's really interesting to hear the unedited version of the conversation because they can hear our side conversations or when we pause to do research, or when we back up to say something again, so they get like a little glimpse behind the curtain. And so we want to provide that to all of you today.

4:31Speaker 1:That's right, and so in the second segment of the show, Matt is going to come on and talk to you about the magic that he works on our episode, and he'll play a couple raw clips so you'll get a peek behind the curtain and see how Daniel and I sound before Matt steps in. And Matt sent a question for us to share with the listeners, and the question was what kind of sound do humans hear best? So let's go ahead and hear what the extraordinaries to say.

5:01Speaker 3:Maybe like the middle sounds like not super high or super low.

5:05Speaker 1:That's what I was gonna say. But the humans are in the middle.

5:08Speaker 2:I don't know what middle of what though, depending on their gender, Like women would hear babies crying the best, because I've heard stories that they often are moiasily working up by that related to our size, Larger animals can hear sounds with longer wavelengths. Smaller animals can hear sounds with shorter wavelength high beach that sounds. Females tend to hear higher frequencies better, possibly to help them hear their offspring when they're crying or something similar.

5:41Speaker 1:Sounds that already stand out from the background and from each other.

5:45Speaker 2:As a new mom, I want to say a baby crying. These are great guesses and hilarious stories. Thank you very much exterdinaries, and they're basically what I would have guessed. You know that there's some range we can hear and obviously ranges we can't hear, and that's going to be tuned to our survival.

6:01Speaker 1:Yes, biology is the answer, and that's great. That's how we know that. It's an interesting question.

6:06Speaker 2:No, it's a harmonious blend of biology and physics. It's biology taking advantage of.

6:11Speaker 1:Physics, biophysics biofirst, but yes, biophysics.

6:17Speaker 2:You know, I think if biophysics means biology, it's modifying physics. So it's at its core physics.

6:23Speaker 1:Well, you know, you and I could be at this all day, and so we thought before we invited Matt onto the show, we would let Daniel do a little bit of his physics magic and talk about the physics of sound. So we're going to dig into that a little bit, and then we are going to bring Matt on. So, Daniel, what is sound?

6:41Speaker 2:Yeah, so sound is not a thing. It's like an arrangement of stuff. It's a traveling pattern of pressure changes in matter. So if something is vibrating like a guitar string, it's going to push and pull on nearby molecules creating regions of compression and the regions when things are less dense, and so those pressure changes then move through the medium. Like if you slap your hand on the table, you're creating momentary high pressure in the molecules on that table, and they push on the ones next to it, and they push on the ones next to it, so that pressure wave travels right. And so this is a wave that needs a medium. It moves through a medium. It's a rearrangement through time of the relationship between the molecules in the medium.

7:28Speaker 1:So I'm feeling even more amazed now by the microphone that's sitting on my desk. How does those like squishiness of the air get transferred into something my microphone can play back to you.

7:43Speaker 2:Yes, so a speaker and a microphone are basically the inverse of each other, right, So let's do a speaker first. A speaker creates the sound. So there's something inside the speaker, like a cone that moves back and forth. It's literally physically mechanically moving at the frequent you want to create. So we think about sound in terms of frequencies the same way we think about light in terms of frequencies you can have like red light and blue light or green light. They are all vibrations of the electromagnetic field at different frequencies. Sound has frequencies also, you can have low frequencies and high frequencies, Daniel frequencies and Kelly frequencies exactly.

8:21Speaker 2:And so a speaker can make these by shaking effectively a drum or a cone at that frequency. So if you shake the cone at four hundred and forty hertz, it's going to create four hundred and forty hertz sound, just the same way that like a radio antenna oscillates electrons at a certain frequency and creates photons or electromagnetic waves at that frequency. So you have mechanical vibration. In the case of sound transmits pressure waves to the air. The microphone is essentially inverse pressure waves in the air vibrates something sensitive, So you're converting pressure waves in the air to physical vibrations of something, and then you just need a divide that's going to record those vibrations. For example, you could transmit them onto vinyl, or you could digitize them and store them on a computer. So a microphone or a speaker essentially inverses of each other.

9:11Speaker 1:Okay, I'm still amazed that we figured out how to do that. That's that is pretty cool.

9:16Speaker 2:It is pretty cool. And there's a lot of really fascinating physics of sound, like how the speed of sound depends on how stiff and dense the material is. So, in air, sound travels about three hundred and forty meters per second. In water, it's faster because water is denser and so molecules are closer together they could push up against each other easier. It's like fifteen hundred meters per second, and stel it's several kilometers per second.

9:41Speaker 3:Wow.

9:41Speaker 2:So sound travels faster through the ground. So you know those scenes where somebody's like putting their ear on the ground to hear if the buffalo are coming, Like, there's real physics there.

9:50Speaker 1:WHOA, that's real.

9:53Speaker 2:That's real exactly. And so the things to understand about sound is that there's frequencies, right, so hind ones and low ones. There's also amplitude, right, is it loud or is it quiet? Those are the crucial things to understand. But musical sounds or voices are much more complicated than just like a single frequency. Right, If you ask somebody with a violin to play a certain note that's going to correspond to a certain frequency, but a violin doesn't produce sound that exactly one frequency. So if you want to play four hundred and forty hertz, then a violin's also going to play a harmonic which means eight hundred and eighty hertz, which sounds like another version of the same note. You know how somebody plays like the low e and a high e and a guitar. They sound nice together because they beat together, they're not shifted, but there are different notes, and so a violin doesn't just play the four to forty. It also plays the eight eighty and the thirteen twenty and the seventeen sixty. And if you play a guitar or a trombone with the same note, it has a different relationship with those harmonics.

10:55Speaker 2:So you can hear a violin is different from a trombone even when they're playing the same note. Why is that Because they have a different set of frequencies that they're playing. It's not a single pure frequency. And this is what we call timber. This is why, like my voice sounds different from somebody else's voice, even if it was at the same frequency, like if you took Kelly's voice and you shifted it down to Daniel range, oh no, it would not sound like Daniels.

11:23Speaker 1:And that this was all new to me like forty minutes ago when we talked to Matt, and so like you know, heads up, Matt is going to mention this a little bit and how he can see that in the software that he uses, and that totally blew my mind. I didn't realize that you were getting like harmonics of people's voices and anyway, very cool, but.

11:40Speaker 2:There's also lots of fascinating biology here, like how the human voice is actually generated. Right, you don't have like a little violin in your throat, but you do have things that vibrate. You have the vocal chords, or sometimes called the vocal folds. They're not like guitar strings. They're more like vibrating valves in an airflow and faster vibration. It's a higher pitch, and so like adult male voices are lower because they have longer, thicker vocal folds that vibrate more slowly. And so this is why also two people sing the same note but sound.

12:11Speaker 1:Different, and birds are somehow able to make at least two noises at the same time, some birds, not all birds, which is just absolutely like if you you should really pay attention to the next time you hear a bird song and try to figure out, like, wait a minute, am I actually hearing that bird make two different notes at the same time? And it's able to do that, and anyway, sound is incredible and.

12:34Speaker 2:Some humans can do that too, right, this is this like Himalayan dual note singing thing.

12:42Speaker 1:Oh, I didn't know that. I know that birds have a very different like voice box than we do, so that they can do that, and so now I need to know what happens with the Himalayan people.

12:53Speaker 2:It's very hard to do, but you can train yourself apparently to do that. And sound is much more complicated than just like generate sound. Here a sound, you can have interference effects because sound is a wave, and so places where the pressure waves are in the same direction will be louder, that's constructive interference, and places where they're pushing in opposite directions it'll be weaker. And this is how like noise cancelation works. Right, you play the opposite sound, the one that's beating down when the other sound is beating up and it cancels it out. So for example, my daughter has air pods in and I say, hey, Hazel, will you take out the trash? And her AirPod plays the opposite of Hey Hazel, will you take out the trash?

13:34Speaker 2:And she hears nothing right, which is why she doesn't take up the trash. Perfectly hypothetical example.

13:41Speaker 1:However, I would like to note that Matt mentions later in the episode that noise canceling headphones don't necessarily cancel out people talking to you. Yeah, I don't want to get Hazel in trouble, but I'm wondering if maybe Hazel's just not listening.

13:54Speaker 2:Yeah, there may be a few effects going on here. But this goes into like the design of concert right, because sound reflects and it can cancel out. So you can have like dead spots in a room where you can't hear someone from across the room, or places where the sound all balances really really well. And this makes a big difference between like a bad concert hall and a good concert hall where like every seat sounds good or more seats sound really good. You could also have beating sounds, like if you have two sounds, they're very similar to each other very close, but not quite on top of each other. Like if you play your guitar tuner at e and then you play your guitar but it's not quite in tune. You'll hear this, Like, those are the waves drifting out of alignment, right, you'll hear them because they're not beating in sync harmonic like for forty and a eighty. They have the down moments at the same time, and so they sound nice.

14:52Speaker 1:I miss that. Could you make that sound again?

14:57Speaker 2:Mac can make that your ring toney for you if you like.

15:01Speaker 1:That would be great.

15:02Speaker 2:And then there's a lot of work that needs to be done when you're like doing noise removal. And Matt's gonna tell us all about that when he gets here.

15:09Speaker 1:All right, so let us delay no longer, actually quick delay we're gonna do. We're gonna do a commercial break, and then when we come back, we will bring Matt on the show.

15:19Speaker 2:The Great Matt Kesselman.

15:41Speaker 1:Matt Kesselman is an audio engineer and producer with over two decades of multidisciplinary experience in the fields of film production, audio post production, and music. Matt has worked with artists and broadcasters like Netflix, the BBC, Lionsgate, Walk Off the Earth, Ill Scarlet and many others. He developed software tools for audio engineers through his company Oslot Audio. He brings over a decade of teaching and communication experience to the table as a college professor at Metalworks Institute, where he is currently and before that at Durham College. He's also the amazing audio engineer and sound designer behind dk EU. He's the guy who makes us sound not horrible. Thanks for joining us, Matt.

16:22Speaker 2:He's also the guy who has a shocking amount of blackmail material on both of us.

16:26Speaker 3:That's right, that's all true. I'm a little nervous. I feel like am I the first guest that doesn't have a PhD? Oh not even in like chemistry.

16:37Speaker 2:Wow, I didn't realize we were gate keeping the guests.

16:39Speaker 1:That's right, Oh my gosh.

16:41Speaker 3:I mean you folks try to find the secrets to life and how to better all of humankind. And I once produced a song called Candy g String. So we're a little different. So I'll try to cold man of the bargain and make this entertaining.

16:55Speaker 2:Well, that's what makes us a harmonious collaboration. Everybody brings their strengths.

16:59Speaker 3:That's right.

17:00Speaker 2:Well, maybe we should start actually by telling the story of how we got connected to you, Matt, how you ended up being the audio engineer for the pod.

17:09Speaker 3:Yeah, how indeed. Well, I was a fan for a long time, and I think, like all the other Extraordinaries, I was listening before we were even called the Extraordinaries. But I always have questions and I always want to know how things work. And a lot of popular science. I can only consume popular science because I am not smart enough, but a lot of it, you know, I don't believe. That leaves me wanting and I feel like there's something missing, and then there's conflicting things in different popular science publications. And then I found you guys, and you break it down and that was really enjoyable. And then I thought, why don't I just tell them that I can make them sound even better?

17:50Speaker 2:Yeah, we got an email out of the blue from you saying, hey, I have notes on your audio engineering, and we dug into it with you, and then eventually we were like, hey, why do you come make the show better?

18:02Speaker 1:And we haven't looked back yet, and so one of the reasons we're super excited to have you on the show is because after Daniel and I do a listener questions episode, we always send the raw audio file export it from riverside out to the listeners so that they have plenty of time to listen to our conversation and send us an audio file responding. And almost every single one of them is like, Wow, this sounds awful. Your audio guy must do so much work, and so we thought that it would be great to have you on the show to like pull back the curtain and talk about like when Kelly forgets to turn off her air conditioner, how do you remove that sound, and like all of the work that you have to do to make us sound good. Where does that start?

18:49Speaker 3:Okay? Well, first of all, I want to point out that there are different kinds of podcasts that have different needs. Right, everybody's familiar with the comedians just talking with no editing, with no nothing, and they leave all the mistakes in and that's part of the charm. They're also narrative podcast that tell a story kind of like a movie, and our very own network has a few examples of that, like Red Elvis and the buzz Aldrin Show. I don't know if you've heard that one, but then there's also the informative kind where you're basically enjoying a lecture. And I feel like this is kind of like a lecture mixed in with some lighthearted comedy. And the thing about lectures is that they take a lot of time to prepare, and you two do something that very few people can do, and that is that you basically deliver two lectures a week, often on topics that you're not specialized in. And it's no surprise that sometimes you need to write on the fly.

19:48Speaker 3:I mean, otherwise, it's just not possible for any person to deliver that much information without having to stop and sort of rewind and think back on it. And I think we all sort of treat it that way, where you have your outline and you have your idea, and sometimes you come up with a new idea while we record, and then you try it out a few times until you get it right, and when you send it off and you hope that I put it together, I feel like.

20:14Speaker 1:You really you get me, Matt, you get me. And I want to start describing the podcast now as lectures with light comedy, because I feel like that's actually that's a pretty good description.

20:25Speaker 3:That's really what it is, right, It's like, yeah, weird things that are very hard to understand, sometimes broken down into small, bite sized, understandable pieces with some pro lapse jokes, which is my favorite type of entertainment.

20:43Speaker 2:We need constant goat updates. You know, people are curious.

20:47Speaker 1:Well, she's kept her outsides on the inside recently, so that's nice, wonderful. Yeah.

20:51Speaker 2:Well, from my perspective, I lean heavily into this editing because I know that we can go back and fix something or say something a better way I speak to friendly than I do in person, or if I'm like lecturing to a class, I'm lecturing to a class. I don't often back up and say things another way and just go with it. So I think the listeners are hearing a very unusual slice of our speaking, you know, the intended for editing but unedited version, which is not how like we have a conversation.

21:18Speaker 3:Normally, Right, And if the listeners want some inside baseball, I always send you the finished episode or a draft of the episode, and then you listen intently and go, oh, actually that detail is technically right, but there's a better way to say it, And then you send me a pickup because you're perfectionists and you want everything to be correct and true, and that's very admirable as well, and you're.

21:39Speaker 1:Always a really good sport when we're like, oh, I actually thank you so much for editing all of that, but I'd like to do it again.

21:46Speaker 2:So let's make this concrete. Can you show us some examples of what you get and then what you send us back. Show us the raw, unedited Daniel and Kelly craziness and then the smooth Matt output.

21:58Speaker 3:Sure, I have two separate kinds of examples. Here's a couple where the noise reduction was already done, and we're going to focus on your writing on the fly, if you will, so take a close listen to. We'll start with with Kelly trying to communicate this idea, backing up a couple of times, a couple of uhs and ums, and then after we'll listen to the cut up version.

22:24Speaker 1:And I have no idea what file met pick. So I'm on the edge of my seat, and we don't really have a good framework yet for predicting. Like if you took you know X, and say you decided you were going to take what's a say you were going to take lions and domesticate them.

22:44Speaker 3:Okay, so there's that. Here's here's the edited version.

22:51Speaker 1:Yeah right, And we don't really have a good framework yet for predicting, like say you were going to take lions and domesticate them. I'm being ridiculous, right, But like say you decided.

22:59Speaker 2:To Really that sounds awesome. I would love to have a house lion. Hey, whatever happened to the house lion I requested?

23:07Speaker 1:Anyway, I need a few more years.

23:10Speaker 3:Yeah. So, I mean you can see that you got this idea of comparing it to lines and it came on the fly, so you just needed to work through it, which is great because I think everybody would rather hear what you intend for them to hear and not for you to sort of back away and just say something that's easier to say.

23:28Speaker 2:And I think that these on the fly moments are what make our conversations really fun because it is a conversation. You're thinking on the fly, you're coming up with new ideas, you're responding to questions. Also, we could never pull off a scripted podcast twice a week. That's just so much more work. And on top of that, I'm a terrible actor, So if I had to read a scriptive podcast, it would sound really stilted. The only way for this to work is for it to be a natural conversation.

23:54Speaker 1:I massively appreciate that I can like stumble and start over many, many, many times and know Mad is going to listen to all of that and pick out the bits that make sense. And you're amazing.

24:05Speaker 3:Thank you. Sometimes you do put a lot of faith in me. And just to be clear, what I mean by that is that you'll have a great first half of an idea, and then you might write on the fly and work out the second half of the idea, and then when I put them together, sometimes they could sound a little bit disjointed, kind of like a milder version of this real segment that really aired on TV.

24:30Speaker 2:An idea, but what we do, we would not be good at what we do, would we We would be sloppy? You calling it sloppy.

24:37Speaker 3:I don't want your recordings to sound like that, even though it's completely normal for speech cadence to change over time, but when you put it all together and remove the gaps in the middle, it becomes more noticeable. So sometimes I have to do things like tune your voice on certain words, or look around and find other times when you said that word in the episode or in other episodes. Wow, use it to help build one cohesive sentence. And just to be clear, I never change what you've said. I might just change how it sounds, but the ideas are purely yours.

25:10Speaker 1:How often does it do we do that to you? I'm feeling really bad right now.

25:15Speaker 3:It's not something you do to me, but it probably happens more often than you think.

25:19Speaker 2:Oh no, no.

25:21Speaker 3:Which is completely normal. You're focused on educating the masses. I'm focused on making it sound smooth without anybody noticing anything was done at all.

25:29Speaker 2:Well, thank you for making this sound so good and making it sound like you haven't done anything. That's the most flattering part.

25:35Speaker 3:You want to hear one of you, Daniel, Yes and no, let's do it. Okay, here we go before and after, For Daniel.

25:43Speaker 2:Had that one piece of data argue that the Earth could be a flat disc that was perfectly that was perfectly aligned to make a circular shadow. Yes, oh boy, that's.

25:55Speaker 3:Real for some reason, like were you at because you hear that in the Listen again.

26:03Speaker 2:And that one piece of data argue that the Earth could be a flat to hear that where Yeah, I was not buying a can of olives or something at the same time as recording the podcast. I've no idea where that comes from. So here's the after Yeah, that one piece of data argue that the Earth could be a flat disc that was perfectly aligned to make a circular shadow. Yes, I sound great afterwards, thank you.

26:28Speaker 3:Yeah, and again, like you said, you don't talk like that in person. It's just that you're trying to communicate things that are new to you in some way and in a way that fits for broadcast.

26:39Speaker 1:Okay, So when I listened to the audio files after I send them to you, I'm like, oh, shoot, I left my air conditioner on. Or Daniel was trying to explain general relativity at the grocery store, like how do you remove like the beeps and the background noises? Like how hard is that?

26:57Speaker 3:So there are many different types of noises in many different solutions. Some are automated, some are not. So if it's here, I have some examples. Here is some unprocessed Daniel, let me know if you can hear that background noise.

27:13Speaker 2:So Bell's paradox involves length contraction, the fact that when things move fast, they look short. So relativity tells us two things, moving clocks run slow and moving objects look short.

27:28Speaker 3:So that kind of noise is probably the easiest because it's consistent.

27:32Speaker 2:It sounds like I'm crinkling paper or something.

27:35Speaker 3:That's a different thing. That's mouth clicks, and that's a normal thing that happens when you put a microphone really close to the human mouth. It's weird how it doesn't happen in person, but it does on micro For that kind of noise, it's because it's consistent. We have software that we basically tell the software look at this, as long as there's no talking, look at this piece of noise, identify it as noise, and from now on remove any time you encounter that. And they be come more and more sophisticated over the years to the point where they're quite transparent at it. It's still in many cases if you go too hard, things start to sound watery and sort of weird, so you have to be careful with that.

28:12Speaker 3:But dig into that a little bit more in detail, because what you're saying is it's consistent, so you can measure it when the person's not speaking, and then you'll have a template to remove it. But still it's varying. So is an average removal. It's a statistical removal because it can't remove the exact noise moment to moment, because there is variation.

28:30Speaker 2:We heard it.

28:31Speaker 3:There is variation, but it's fairly predictable. Like an air conditioner, unless the air conditioner is changing temperatures, is going to give or take make the same droney noise throughout the recording. But yes, sometimes I do catch that throughout the recording. Sometimes the air conditioner change gears and then it's no longer being recognized and I have to compensate for that. There are settings where it's supposed to follow the noise itself.

29:01Speaker 3:I don't find those to be very good yet, but tuch changes all the time. Maybe tomorrow there'll be a good one. And is this noise removable? Because it's at a different frequency than the voice, so you can identify it when I'm not speaking, and then you know exactly how to remove it. And you can remove it without removing the voice or does it overlap of the voice and so you are affecting the actual voice as well? Oh okay, that's a great question, you really want to look under the hood. So the way it really works is that the recording is divided into smaller bands of frequencies, and each frequency is given what's called a gate or an expander, where if a lot of volume hits that area, it's allowed to pass through, so speech, for example. But if there's a quieter moment, it enhances the quiet moment and makes it even quieter. So if you really listen closely, you may be able to notice that there are moments of noise when you talk. But because you're talking and then noise is only present as you're talking, it's not perceptible.

30:04Speaker 3:A lot of this is about magic tricks and not actually about doing what you think we're doing.

30:08Speaker 2:By magic tricks, you mean distracting people and understanding how works.

30:12Speaker 3:Yeah, I mean I mean real real magic, not magic magic were slide of hand and stuff like that.

30:19Speaker 2:I love what you call that real magic. That's awesome.

30:22Speaker 1:Okay, So then how would you remove something like the grocery store.

30:26Speaker 2:I was not at a grocery store. I want that on their front.

30:29Speaker 3:I think you were.

30:30Speaker 1:I saw it.

30:31Speaker 3:I saw it. So for a noise like that or like goats, Kelly oh yeah, their goats. Yeah.

30:42Speaker 1:Oh, I'm so sorry.

30:43Speaker 3:Oh no, it's totally it's not your fault that they're goats. They do goaty things, uh, for things like that. For for noises that are not consistent or predictable, which, by the way, let me back up one second. Noise canceling headphones work on the same principle, So noise can headphones can cancel predictable noise once they recognize that noise. However, if somebody's talking to you or there's a dog barking, you may notice that your noise canceling headphones don't work for this very same reason. So if there is a dog barking or a goat. Luckily, we live at a time where I can just open it up in a spectrol graph and I can actually see all the sound.

31:21Speaker 3:It looks like splotches, and I can see the goat bleed to an extent, and I can dry it out like an photoshop.

31:28Speaker 2:WHOA, So what you're looking at is the amplitude versus frequency.

31:34Speaker 3:Yeah, time is the x frequency is the y axis, and then amplitude is color. Oh bright orange means it's very loud. So I can actually see all the harmonics of the goat or a door being slammed, or a plane flying by, And sometimes it's hard to draw them out because there's other stuff happening there. But in most cases it just takes time. You have to stop, capture that piece and follow the goat and remove it.

32:01Speaker 1:How does Daniel's voice show up different than mine? Go So Daniel has a very deep voice. What does his look like relative to mine?

32:10Speaker 3:Well, too bad. This is not a visual show. So what you would see for any human is just a series of lines on top of each other most sounds, really, and the lowest line is the actual note that Daniel or Kelly are saying. And then all the lines the notes above that are harmonics that we don't perceive as individual notes, but they're the sort of fingerprint that make Daniel and Kelly sound different. Even if you sing the same note, for example, is just a different combination of those harmonics, huh.

32:42Speaker 2:Like, for the same reason that like, a guitar and a violin sound different when they're playing the same note, because there's a range of frequencies that gives you the mental fingerprint to identify it exactly.

32:53Speaker 1:That's all it is, all right, Well, let's take a break, and when we get back, we're going to chat with Matt about how software is making his life easier or maybe not.

33:02Speaker 2:And whether or not he's actually an AI.

33:04Speaker 1:And we're back and we've got Mat in the hot seat today we are quizzing him about how he makes us sound not horrible, because we sound pretty bad in the raw audio that we send him at So AI is like everywhere now, so could you like tell AI? Hey? Here is an interview from two people for Starters. Cut out every time they say and remove all the background noise like are we there where that could happen? Or yeah, where are we with AI?

33:49Speaker 3:Now? Uh, We're not there. We might be there. I have no idea what's going to happen. Maybe next month none of us will have jobs. I know this. People have very strong beliefs about what's going to happen. I think nobody knows, and we may hit a wall or it may never stop. See. The thing is, there are tools that can do things like that, but then they require adult supervision where you check whether it did and then you find that it deleted a very important question or you know, things like that. So it doesn't help me It's still faster for me to do it all myself because I work pretty fast, and you know, I don't have to go back and try to find what's missing.

34:29Speaker 3:The tools I use the most with AI are for noise cleanup. That's also another option if I don't draw it out with a pencil tool. Is There are AIS that are pretty good at being able to tell what noise is, and they sort of split into two categories. One is it removes the noise on its own using all kinds of cool phasing algorithms, and then there's one that recognizes your voice and then resynthesizes your voice completely from scratch. Wow wow, which is a little weird, scary scary. Yeah.

35:04Speaker 2:Well, I had a listener wright to me once that they heard an AI generated podcast and that one of the hosts on that podcast sounded a lot like me, and I was like what, So then I listened to it and I was like, that does kind of sound like me, And then I realized there are hundreds of hours of my voice out there for anybody to train an AI to make an AI Daniel.

35:25Speaker 3:And you only need five seconds now. And by the way, PSA, I don't know about your bank but my bank has this voice print identification if you call them, Yeah, I asked them to turn that off. Yeah, because it's incredibly easy, especially for anyone that has their voice out there, for anybody to fake your voice. So that is no longer safe if you have that enabled at your bank.

35:46Speaker 2:And I'm not endorsing it, but I did find this AI generated podcast to be pretty good. And I did a little experiment where I took one of Katrina's latest papers and I ran it through and I had produced a fifteen minute podcast about it. Then I sent it to her and didn't tell her it was AI generated. I said, hey, I heard this podcast about your latest paper. And she listened to it. She was really excited and she shared it with her laugh.

36:10Speaker 3:Huh, that is weird.

36:11Speaker 2:It was surprisingly good. It's pretty compelling stuff.

36:14Speaker 1:Unfortunately, Well, so, I think you should be flattered that out of all the voices they could have picked, they decided to scrape your voice. Daniels a good voice.

36:21Speaker 2:I don't know that they did, you know, just generic podcast dude with a deep voice, I guess. But no AI is out there making goat jokes, Kelly, So you are unique?

36:32Speaker 1:Yes, Oh that's great. I'm so glad, So Matt. Even with the help of software, we send you about an hour, an hour or twenty minutes of audio, it gets compressed to about an hour of podcast. How many hours go into editing? Uh?

36:50Speaker 3:Anywhere from five to ten wow, depending depending on physics or bio subject matter. Yes, yes, Well, look if it's if it's about particle accelerators, Daniel already has all that stuff memorized and it's easy to do. Or if we're talking about killie fish or something like that. But if somebody just throws you a curveball question that you had to look up, then it could take a bit longer. If there's a guest, it usually takes longer because you have a rapport and a banter that you're used to, and then if somebody else comes in, they may not have that same vibe and it takes a bit of time.

37:28Speaker 3:And I want to point out that five to ten hours may sound like a lot, but it's not a lot or unusual at all if you think about it. It's usually roughly about an hour and a half of raw audio that I get. I have to listen to it at least one time, so that's already an hour and a half. Then I need to do some noise reduction, which can take thirty to fifty minutes, depending on the type of noise there is. And then from there, every edit that I make, I have to first hear it, stop, listen to it one more time to make sure there is a problem, do the edit, which takes between a few seconds and a few minutes depending on the edit, and then listen back. And these little things all add up, and before you know it, you're easily crossing the five hour mark. By the way, one other thing that I forgot to mention is there's the editing sort of to help everything feel smooth and makes sense. But then there's also the issue of comedic timing over web conferencing. There's a delay, whether you like it or not, and it's not something as easy as shifting one voice back. Because Kelly would say something funny, Daniel would accidentally talk over

38:35Speaker 3:Kelly because he's been waiting because there's a delay, then realize that was funny, laugh at that. Kelly also hears a delay, so she's already trying to damage control Daniel not laughing, and it becomes this mess where I'm like, there's a very funny joke here, and both laughed at the joke, but because of that delay, it sounds a little bit train wreckage. So that's another thing that.

39:01Speaker 2:I know, Wow, what a review?

39:04Speaker 3:Well, because I know that in person you both would have this banter and laugh and it would be so I'm sort of time traveling in a way to put you where you wanted to be, at least where I perceive you want it to be, so that that's also part of it.

39:19Speaker 1:Thank you for making us funny.

39:21Speaker 3:No, you're the funny ones. I just undo the damage the internet causes.

39:26Speaker 1:So why does my voice sound so much better in my head? So like when I listened to the file you send me, I think, wow, this sounds way better than when I actually was like talking to Daniel in this version, Daniel left at my joke right away, but I still I still cringe because I don't love the sound of my voice when I hear it back to me. But when I'm saying it, it doesn't sound so bad. So why is there that disconnect?

39:51Speaker 2:I think you have a great voice, Kelly. I love hearing it.

39:53Speaker 3:Thanks Daniel the great great voice.

39:55Speaker 1:Thanks guys.

39:56Speaker 3:Otherwise people wouldn't be listening as interesting as you are. If you had bad voice, as people would move on. It's a matter of perspective. So we hear everything else and everyone else through our ear canals. Right, there's air outside, and that air vibrates and that goes into our ear canals. Will hear it for ourselves. A big chunk of our voice travels through conduction into our ear from the inside, and that's what people call the inner ear. There's no actual inner ear. I used to think there was when I was a kid, like a third ear in your throat.

40:25Speaker 3:There isn't one. You're basically hearing sort of a muffled version of yourself because it's traveling through soft tissue and bone, and then you're also hearing a little bit of yourself through the air going back into your ear, so it's a different sort of like I don't know. When I look in the mirror, I'm gorgeous. When I look at a photo, I look like a crescent moon. I noticed that my face is curved, and I never noticed that in the mirror because it's just a perspective I'm not used to. And I think that's exactly what happens with your voices. You're used to hearing it a certain way, and no matter how the recording sounds, you'll think it sounds worse because you're just not used to it.

41:04Speaker 2:So somebody was to put their cheek up against Kelly while she was speaking, so that her voice traveled like through their soft tissue, would they hear Kelly's internal version of her voice?

41:14Speaker 3:If you were somehow able to model Kelly's innards from her throat all the way up to the top of her skull and put that between the other person, yes, it would be very strange.

41:28Speaker 1:But yes, I'd just like to put on record that when I see you, I never think Crescent.

41:33Speaker 3:Moved, but I will now I have to tell myself the same thing is that that tiny lack of symmetry which every human has is just accentuated because I never see it from that angle. And I feel like most people when they look at a photo compared to the mirror, they're like, oh, what is that? What happened?

41:50Speaker 1:Yeah?

41:51Speaker 2:So you mentioned earlier about real magic and how big part of this is knowing how it's going to be received by people. Tell us more about that, about how understanding the way sound is actually perceived and the way people pay attention changes how you do your editing and tell us about psychoacoustics.

42:09Speaker 3:So, Daniel, you're gonna be proud of me. Before I even knew who you were. I've talked to my students about aliens a.

42:16Speaker 2:Lot plus ten points.

42:21Speaker 3:Yeah. The reason I do that is I always in the first class. I tell them this is a class for human beings in Earth's atmosphere, because, first of all, sound behaves differently through different materials, and second, humans have evolved to process their senses in ways that are useful to them, which is another way of saying that what we hear and see and feel is not reality, but our useful version of reality, and sound is no exception to that. I think we may not realize how much of a role it plays. My theory is because there are no earlids go on. With sight, you have these eyelids where you can close them and not see and get an idea of what it's like to not have that sense. You also have to directly look at the thing that you're looking at to perceive it, and with sound, you hear everything around you. You don't have to look at it with your ear, and you've never closed your ears.

43:24Speaker 3:Your earballs are always exposed. So it's a sense that's so enveloping and so much part of our lives that we sort of take it for granted and not understand how much information we're constantly getting from our surroundings through sound.

43:38Speaker 2:And we don't even have the language for it, right, you have to adapt the language from eyes. Like you said, look at it with your ear instead of hear at it with your ear, because that makes no sense, right.

43:49Speaker 3:Yeah, So we think there's just this baseline of nothing, but in reality, you know the size of the room you're in and what the walls are made out of before you look and tap them. Walls, you know from how sound reflects in that room.

44:03Speaker 2:I have a question about that because I've been told many times to record in a room that doesn't have auditory reflections, right, like surround yourself with laundry or something, because nobody wants to hear the room that you're in. So isn't that weird for people to hear audio that sounds sort of like disembodied or de rumified.

44:22Speaker 3:No, because it's far from disembodied and de rumified unless you've been to an anichoid chamber. You have never heard silence in your life. It's a have you been to one?

44:31Speaker 1:No?

44:31Speaker 3:I highly recommend it. It's weird.

44:34Speaker 2:Sounds like a torture device, you know what.

44:36Speaker 3:People say that there's a rumor that no one has ever lasted more than forty five minutes. Not true. I've been there for hours. But you can hear blood rushing through your ears. You can hear everything. And not only that, if you're there with a buddy and they turn around and talk and speak the other way, you can barely hear them. We don't realize, wow, or we take for granted how much of what we hear is actually reflections of the source and not the source itself. So I mean an example of some important psychoacoustics that engineers have to keep in mind. And also the answer to the listener question, what do you think what is the type of sound or sound that humans hear?

45:16Speaker 2:Best? I would have guessed that there's something similar to our visual arrange, where there's a whole spectrum of frequencies and we can see something in the middle. Where in the middle basically means where we can hear, which makes it kind of a non.

45:28Speaker 3:Answer, right, Well, that's actually that's a good comparison with sound. You can see red just as well as you can see blue with sight, but with sound it doesn't work that way. With sound, we can hear frequencies from twenty vibrations per second twenty hurts to twenty thousand vibrations per second twenty killer hurts, but out of those we don't hear them all equally. The frequency range we hear best is between one and five killer hurts, and that is the range of leaves rustling and breaking.

46:00Speaker 2:And jaguar is creeping up on you so that.

46:03Speaker 3:Predators don't creep up on you. It's also the range of a baby crying. You can hear a baby cry from way farther away than I don't know a car engine reving at the same distance, same volume.

46:15Speaker 2:It's also the sound of a candy bar wrapper, isn't it That is?

46:19Speaker 3:Yes, yes, it is, and we've been using that. For example. Police sirens are tuned to that frequency range, and that's why you may have found yourself sitting in your car or walking hearing a siren looking around being like, I don't see the police anywhere. Where are they? It's because we're so good at hearing those frequencies that we can hear them from so far away that we can't even see the source of the sound yet.

46:40Speaker 1:Wow.

46:41Speaker 2:So then how does all of this shape your editing? Are you trying to push my voice up to police siren frequencies?

46:48Speaker 3:Well, I definitely have to make sure that intelligibility is there and that is in those frequencies. But there's a challenge here because we hear those frequencies so well, they become annoying very quickly too, because we're acutely tuned to them. So there's a balancing act where if there's not enough it sounds I can show you. Do you want me to show you? Yeah?

47:11Speaker 1:Yeah, please?

47:12Speaker 3:So here let me play process Daniel here.

47:15Speaker 2:So Bell's paradox involves length contraction, the fact that when things move fast, they look short. So relativity tells us two things, moving clocks run slow and moving objects look short.

47:29Speaker 3:So now I'm going to take out that frequency range that I was just talking about, and you'll notice so you can still hear Daniel. He's loud, but the part that makes him human loud, I mean loud is the volume I made him. Not because Daniel is a loud person, but the intelligibility is just not there. Take a listen.

47:47Speaker 2:So Bell's paradox involves length contraction, the fact that when things move fast, they look short. So relativity tells us two things, moving clocks run slow and moving objects look short.

48:01Speaker 1:Huh.

48:02Speaker 3:And now I'll exaggerate those frequencies and you'll see that it'll kind of hurt your ears.

48:06Speaker 2:So Bell's paradox involves length contraction, the fact that when things move fast they look short.

48:13Speaker 3:Ooh right, So I have to find a happy medium, and that is sometimes something that has to be done temporally. So I have to make sure that some words don't have too much of those frequencies and others do. That's just one of many ways in which I have to look at the waveforms and understand the science and then also understand the psychoacoustics, the psychology of how humans perceive this stuff.

48:36Speaker 2:Well, you're doing a great job, because I sometimes meet listeners in real life and they'll say to me, oh my gosh, it's so weird. You sound just like you do on the podcast. So you are reproducing this faithfully.

48:47Speaker 3:I did it, woo.

48:48Speaker 1:So how much physics did you have to learn to be this good? At your job.

48:53Speaker 3:I see what you're saying. Yes, audio engineers are real engineers. We're like the dentists of the audio profession.

49:02Speaker 2:She's just trying to sus that if you're on the physics camp or the biology camp, she's wondering if you're biased.

49:07Speaker 3:See, I'm stuck right in the middle between physics and biology exactly.

49:11Speaker 1:It's a good place to be.

49:12Speaker 3:Yeah, it's it's a very fun place to be because who cares about physics if it's not for how it affects sentient beings?

49:20Speaker 1:Amen?

49:21Speaker 2:Amen, hmmm, I have thoughts of that.

49:24Speaker 1:I'm so glad we got you. You know, Matt, you should be on the show every week.

49:28Speaker 2:I see, Kelly just wants to invite people on the show who will disagree with me gang up on me.

49:35Speaker 3:So let me tell you another psycho acoustic phenomenon called masking. The way this works and this probably like we can't ask evolution, right, but it looks like we've developed this property to u curb over stimulation. What happens is if there are certain and it's very predictable, if there are certain frequencies present, we are deaf to certain other frequencies in those moments. It's just like a blind spot in the car, Like, no matter what mirror you look at, you will not see the car right next to you if they're in your blind spot. So we have these blind spots that are constantly, very quickly changing as different sounds are present, and very smart programmers mapped out the exact way we have these blind spots and figured out, you know what, we can save a lot of space and audio files if we just throw that stuff away.

50:28Speaker 3:Oh oh wow. And that is how the MP three file format was born. It does throw away a lot of things, but it throws away only things you would not perceive anyway. And so if you've ever had this debate with some or just heard someone claim that MP three's just don't sound like waves, man, there's just something missing. Not necessarily, if it was a well encoded MP three, you will never hear what is missing. Wow, Because you're human. Our imperfect hearing has really helped us on file sizes. However, if an alien came and you went, here's a wavefile, like here's a recording, and then here's my MP three of that recording to that alien, it might sound completely different and upsetting because they don't have our same perception.

51:14Speaker 2:Those aliens who listen to the podcast might be really sensitive to those frequencies that our audience, our human audience, can't hear.

51:22Speaker 3:As far as we know, when we listen to an MP three, our pets are suffering. They don't seem to be complaining, but they're definitely not hearing what we hear when we listen to the original recording the way that humans do.

51:33Speaker 2:So this is like the moral equivalent of masking out the ultraviolet off of a painting, and a bee might see a painting very very differently, whereas a human would be totally unaware.

51:44Speaker 3:Kind of It's like if humans, when you see yellow, you're temporarily blind to purple.

51:51Speaker 1:Weird, Why why why does that happen?

51:56Speaker 3:I don't know. I haven't asked God yet, but I think probably because of the overstimulation thing. I think if we were to hear everything all at once, our brains would explode. I think it's just too much information.

52:09Speaker 1:So people who get over stimulated, do they hear some of that or is that just a different thing?

52:15Speaker 3:You know, that's a great question. That might be part of it, That might be part of it that they don't experience masking the same way. But the interesting is that it's very predictable, this masking and how it happens. It's very specific, and it's the same for children and for women and for men and for everybody. I don't know that there's a group of people to whom MP three's don't sound right objectively, like a blind test. Yeah, so yeah, I don't know. Maybe some people do experience less masking. Maybe that's part of it.

52:47Speaker 1:So, Matt, I imagine you have loads of fun stories, in part because like Daniel and I always hit record and then we're like, we totally forgot. We hit record, and we have conversations about the most personal things happening in our lives, and then we send it to you, and so you it's a very asymmetrical relationship. I think you know a lot about the things that are going wrong in my life. I don't know nearly as much about what's going on in your life.

53:09Speaker 3:I can make you a list if you want to even the playing field.

53:13Speaker 1:I mean that lay it on me. But but so I imagine that one you know a lot of secrets from people and tell us some weird, funny stories from from life as an audio engineer.

53:24Speaker 3:Uh, there are many inappropriate stories. Uh there's okay, here's one that is right in the middle between science and and some biology and perception, so mostly physics. So I was in the Santa Monica, which is uh is it a part of LA or is it outside of La boy Daniel, California.

53:48Speaker 2:You should know this. I don't know. It's well, you know, there's the affective LA which is a big blob of urbanized area, then there's La County, then there's LA City, and I don't know the distinctions between all of them.

53:58Speaker 3:Yeah, I don't understand how that's a municipality because when I was working there, they were shuttle us around from location to location, and the names of the neighborhoods and how things fit into the municipalities didn't make any sense. Anyway, I was in Santa Monica filming with Smokey Robinson, Wow for a very interesting documentary by the way, called Paid in Full on the BBC and CBC wherever you are, about the unfair treatment of black musicians throughout history. In any case, I was there with Smokey Robinson and they rented a special villa just for him. You know. The VIP treatment, and there's always this air of when a really important person comes in, no one stops the recording for any reason. Right, do your job be quiet. We have a very limited amount of time with this person.

54:53Speaker 3:So I'm sitting there, I'm set up, I have my computer open, I've miked him up, I have a mic on him, I have a mic above him, and I have a spectrum analyzer open, just because I didn't really need it, but it shows you all the frequencies in real time. And suddenly during the interview, I see these sort of spikes at like five hurts, which is way lower than what humans can hear. And I'm realizing that even though the microphone is suspended and they're supposed to not pick anything up through the floor, I've created a seismometer by accident. Oh and I can see that there's an earthquake coming.

55:30Speaker 2:Oh wow wow.

55:31Speaker 3:And so in the middle of this interview, I go, hey, excuse me, and the director looks at me, like what are you doing, And I'm like, I'm very very sorry. I think there's an earthquake coming and we should get ready to evacuate or move away from the walls or get too though I don't remember what you're supposed to do. But we're in California. I expected everyone to know except for me, and they're like, what are you talking about? There's nothing happening, you've this is Smokey Robinson. What are you doing? I mean, everybody was nice, we're all friends.

56:02Speaker 3:But as that's happening, the lights start to shake. Oh wow, and there's a rumble. And luckily it wasn't a big earthquake or anything. Everything was fine. But I predicted on an earthquake using a microphone and a computer screen.

56:16Speaker 1:That's amazing.

56:17Speaker 3:And I interrupted an important recording.

56:19Speaker 1:And that's magic, right, it's Smokey Robinson is now like Matt Kesselman is magic. Everybody knows.

56:24Speaker 3:He was like, that's okay, baby, he's got a very very soft voice.

56:27Speaker 2:I love him.

56:29Speaker 1:Well, that is a great story. I don't have any comparable stories for warning famous people about impending natural disasters.

56:37Speaker 3:It's a very niche story.

56:38Speaker 2:Market all right, well, thank you for coming on the show, and we're telling us all about what happens in the background to make us sound so good. Before you go, tell people where they can find out more about you and what's.

56:51Speaker 3:Up to you. Can find me for everything at Matt Kesselman dot com. That's m A T K E S E l m N dot com. And if you don't like to spell things that are complicated, I just bought Thegoshfather dot com and dot cost to my website too, so you can use that instead.

57:13Speaker 1:Matt, I appreciate you so much for so many reasons, and now I appreciate that you came on the show. You're the best.

57:18Speaker 2:There are lots of reasons why the show sounds good and it's fun to listen to and easy to listen to, and many of those are because of Matt. Thank you very much.

57:26Speaker 3:Amen, thank you for having me. This was a lot of fun.

57:35Speaker 2:Thanks everybody for listening. Please go and do us a favor and rate the show on whatever podcast app you're using. It really helps people find us.

57:43Speaker 1:Daniel and Kelly's Extraordinary Universe is edited by the amazing Matt Kesselman.

57:48Speaker 2:He really is a wizard. You can also find us online on Blue Sky, Instagram, and x d K Universe. Come engage with us.

57:57Speaker 1:You can email us I had questions at Dan and Kelly dot org. We really do want to hear from you, and you.

58:03Speaker 2:Can find our website www dot danieland Kelly dot org, where you'll also find an invitation to join our discord where everybody comes and talks about the amazing.

58:13Speaker 1:Universe, and we also have the most amazing moderators. This is an iHeart podcast. Thanks for joining us.