Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
I don't feel like working today, Laurel.
(00:04):
Oh, let's talk.
Oh, all right, we should talk instead.
Okay, what's today's date?
I don't even know what today's date is.
Today is September 27th, that's a Friday.
September 27th.
I saw you looking at your watch to know what day it was,
which is a total normal thing,
unless you have a dumb face like I do on my watch.
(00:26):
It does not tell me what the date is.
That's very sad.
Yeah, I cannot find a way to,
anyway, that's not a huge problem,
and it's certainly not a creative problem, but.
Welcome back to our podcast.
And we wanna talk about creative issues
(00:47):
and things that people who are trying to be creative
and make a living out of their creativity
or that creativity are facing,
like speaking into a microphone.
And well, we don't, this was totally impromptu
because yeah, we were setting up our audio
(01:07):
and decided why not.
We should talk audio.
You've been looking at stuff
and I've been looking at something completely unrelated
with AI.
Oh, really?
What have you been looking at?
That's interesting.
Well, one of the things that we're trying to do
is eventually get our audio books,
(01:30):
well, get our books as audio books.
Yes.
But a lot of the low budget authors
are starting to use AI readers for their audio books.
I know, that's horrible.
Please don't say that we're going to do that.
No, no, we would never do that.
I think that even I can do better
than the automated readers,
(01:50):
but that it is becoming more pervasive.
And it's something, yet another thing
that threatens the livelihood of creatives.
Yeah, because you have the readers
who make a living from reading books.
But boy, I don't know what it is.
That just sounds awful.
Every time I go, if you're a YouTuber
(02:11):
and you're considering this,
maybe think about it carefully
because every time I see a video and I start it
and the voice is fake, I immediately turn it off.
Yeah, me too.
I just, I can't stand that fake voice.
And the other day, I think we were looking
(02:31):
at some audio things
and I was looking at a professional software
that costs $1,000
and the tutorials on and from the company
are with a fake voice.
And I'm like, I can't do that,
especially for audio software.
(02:52):
Anyway, that's my rant there.
But that's interesting.
So they're gonna be improving.
They are gonna be improving.
The other thing that's interesting that's related
is accessibility, right?
In one of the social media that I do called Substack,
a lot of the articles are longer.
It's a lot of the people on there are writers
(03:15):
and they're starting to have automatically
as part of the thing for an AI to read your sub stack.
So if somebody has either vision problems or.
Oh, that's good then.
Well, but you could also do it yourself.
So then the argument comes up,
you're starting to see people say,
(03:35):
hey, they're starting to put AI on everything.
If you don't want a robot reading your words,
then do it in your own voice.
Oh, interesting.
So you can, so let me get this straight.
So you can choose to have your own voice for what?
For like articles that you write.
Right.
Oh, and you can like say,
(03:57):
okay, here's the audio version of this article.
Exactly.
Oh, wow.
Okay.
Or if you don't do anything.
That makes sense.
Then the platform is going to have AI read it out
so that people who have accessibility issues
can still participate.
Wow.
So I'm blanking out on the name of the actor.
(04:18):
So please say it as soon as I tell you.
But the voice of Darth Vader.
James Earl Jones.
James Earl Jones.
And he passed away recently.
Now imagine, because he did have a contract with Disney
where he was willing to donate his voice after his death
(04:40):
to do Darth Vader's voice.
And they recorded a lot of samples, they said, of his voice
so that together with some editing and AI,
the AI could reproduce his voice.
And so, you know, and I haven't read like all the details
about how he decided to do this.
(05:04):
I know that was part of the contract negotiations
last time SAG was, you know, going in and renegotiating
because they wanted to make sure
that there was fair compensation,
that you didn't like, you know, sign off once,
pay a small fee, and then you own their voice
or their likeness forever.
But then imagine, okay, James Earl Jones,
(05:28):
reading your sub stack things with the AI
and that let's just say that we are a bit into the future
and it sounds just like him reading your novel,
reading your voice.
Then what?
Well, that is the interesting thing
because it's never, you know,
it's never the same as a real actor.
(05:49):
There's a reason that they prepared for their roles, right?
It's more than just the tone of their voice.
It's the way you choose to emphasize words.
It's the way, you know, an AI can kind of pick up on,
oh, there's an exclamation point, you know,
or a question mark, but it can't do shades of difference.
(06:14):
And especially when it comes to making different characters
sound a lot different, there's a limit
to what an AI can figure out
versus what a human reader can put into the story.
And that's interesting.
So software right now, AI does not have,
(06:40):
let's call it a soul, right?
It doesn't, what is it not good at?
Humans are better at feelings than AI.
And if you're in a creative field, you know,
and we've certainly learned that feelings are, you know,
a big part of why anybody likes art, you know, like it's,
(07:04):
underestimated a lot of times,
and logic is usually put above, but sometimes people do,
you know, especially by art out of feelings and emotions.
And so we're talking about the idea that this AI,
as good of a reader as it may be at faking it,
(07:24):
it's going to follow some logical steps
for this is this word, I'm gonna say it like this.
And yet there are a lot of people who still buy records,
you know, and I think about the idea that people look
for imperfections in a way, you know,
some type of randomness, you know, like, even in audio,
(07:47):
we know that people like to add warmth, you know,
which we've learned is some distortion in a way.
And they like that aspect of it.
So do you think that it will ever replace, you know,
and it's hard to say that, you know,
like will it ever replace humans doing that?
(08:08):
But I hope not.
I hope that, you know, I mean, if they can read the stories,
they certainly can already write the stories.
I've got to hope that we will continue to see the value
in the difference between what an AI can create
(08:28):
or produce rather than what we ourselves can do.
Because, I mean, you're talking about the, yeah,
the imperfections.
As I was listening to it, I didn't realize at first,
I hadn't looked at the title details
to see that it was an artificial reader.
(08:50):
And I was thinking, you know, either this reader is like,
not into the book, right?
It sounded like they were bored or something.
You know, a little, I mean, it was good enough,
but there was no sense of investment in it, you know?
Yes.
(09:10):
And it was all kind of exactly even.
And I thought, you know, I wonder if this is one
of the books that's read by an artificial reader.
And when I went back and checked, I'm like,
that's totally it.
And it's funny, because there isn't something
you can really point at.
It's not like, they're not as bad as the ones
that read my textbooks for school.
(09:31):
The ones that read the textbooks for school are awful.
You would never ever confuse it for a human.
Or even a really decent robot.
Right, right, right, right.
But there's still this level of evenness
that you just, even the imperfections in the human reader
(09:55):
add to the feeling that, you know, that the book is real,
that the voices are real.
Yeah.
And, you know, I could see how eventually they might think,
oh, we're gonna introduce some randomness
just to make it seem like, you know,
(10:16):
a little bit more imperfect and stuff like that.
But even that, you know, I'm a software developer
and I know that even that randomness is programmed in.
You know, it's not like completely random.
There's going to be some pattern to it
that after a while your brain is going to be able
(10:38):
to recognize.
So, but I can imagine, for example, you know,
weird voicing our own characters.
And I, one of our characters,
it's a bacon wrapped avocado mummy, you know,
blah, ha, ha, is what we call him.
And my voice for that is a really bad impersonation
(11:00):
of Stallone, you know, Sylvester Stallone
doing his rocky voice.
So it's so bad, you know, that people can't tell
that that's what I'm trying to.
Which is great, it makes it unique.
To imitate, but that's, you know,
that's what I think about when I'm doing the voice.
(11:23):
And, you know, I can imagine, sure, you know,
even now probably, if I go online and search for Stallone AI,
there is something that can, you know,
do it better than I can.
But hopefully my acting and the fact that it's a bad
(11:45):
impersonation and all this stuff.
Adds to the humor.
Adds to the humor, adds to the character,
gives them a real personality that AI, you know,
should not.
But I can totally imagine in the future,
you just go to the computer and say, okay,
this character is going to sound like this and like that.
But I really hope that our brains are capable
(12:10):
of detecting that, you know.
Well, it was interesting because I think it was yesterday
that we were listening to the meta announcements.
Yeah.
And they were talking about, you know,
now for their echoes and some of their AI,
you can pick voices from real actors.
And the demonstration was using Awkwafina's voice.
And it was close enough that you could tell
(12:32):
that it was supposed to be Awkwafina.
But it lacked any of the real humor, you know.
Yeah, interesting.
I mean, I love her voice.
I love her acting.
I love her characters.
And I thought, oh, that would be kind of funny
to have that kind of, you know, snark coming out
of our echoes.
And it...
(12:54):
It's not going to be her, right?
It's not going to be her, you know.
It gives you the hint of, so sort of a reminder of,
but not the heart of who she is and what it's like to hear.
And that's the thing that I think we are safe
as artists for a while because the machines cannot,
(13:17):
do not have emotions.
You know, right now the machines cannot feel on their own.
And I feel like that, you know, not to overuse the word,
but I feel like that thing in us and humans
that produces our feelings and produces our personality
and all that stuff cannot be replicated yet.
(13:41):
And it's essential for people to love art, you know.
I just imagine like an AI painting
and doing paintings and being in a museum.
And they might look perfect in some way, you know,
and you might even be able to say,
(14:02):
hey, I want you to imitate this, you know.
I want you to come up with the Mona Lisa, you know.
And I have to believe that our brains will still be able
to look at the real thing and look at something produced
by AI and say, oh, you know, that's, that's,
(14:23):
this is the real one because it makes,
it evokes an emotion in me that the other one doesn't.
You know.
Yeah, I think, I think that is something
that you have to hope for.
It's not just the expressing of emotion,
but that connection that you get between the artist
and the consumer where the person looking at the art
(14:48):
or experiencing the art can connect with that emotion
and connect with the part of themselves
that feels that emotion that I don't think, you know,
I would be very sad to find out that we started having
(15:08):
that connection with something made by a computer
without, you know, realizing it.
We are already so separated from each other in many ways
and people talk about it.
It would be sad if we started sort of feeding
that loneliness with computers rather than with people.
(15:30):
Right, right.
And I don't know about you, but sometimes you go,
people share images on Facebook or Instagram or whatever.
Can you tell or have you encountered this
where you can sort of tell, oh, was this generated by AI?
(15:53):
I'm not normally in spaces where that would be,
where AI imagery, most of the spaces where I'm at,
people are creators and so they're very against AI imagery.
I just encountered this in Facebook where one of the people
who are my friends in Facebook was sharing some images.
(16:16):
One was a cat dressed up as a cowboy
or something like that.
And it had this like, I don't even know how to call it,
but there's something about it that I could tell
this looks like, you know what it was?
Oh, I know, I'm gonna date myself saying this.
(16:41):
Glamour shots.
Oh.
Do you remember?
Okay, if you're old enough to remember this,
there was, this used to be in the mall, I think.
Used to be, in lots of malls.
In lots of malls.
You could go to these places
and get what was called a glamour shot.
And that was a picture that was supposed
(17:02):
to make you look better.
But whenever you saw a glamour shot picture,
you could usually tell that it was glamour shots, right?
Something about the filters that you used.
Something about the filters.
Something about the way, yeah, it just looked,
I don't know if the word is too perfect,
but it was just soft and it had that look like
(17:26):
it was sculptured, you know,
and all your blemishes were gone and all that stuff.
Processed.
It was heavily processed and you could sort of tell.
And right now, anyway, AI images have that type of filter
and feel to them, you know?
(17:47):
And sure enough, somebody in the comments said,
now is this a generated images?
And the person, you know, that I follow, my friend said,
yeah, I've been like just entering things in AI
and posting the pictures.
And you know, and I kind of wonder, you know,
(18:07):
how soon before people go, hey, can you stop that?
Or, you know, because some of them are funny, you know?
But something about it being generated
takes away from the humor and it's like.
It almost feels dishonest.
Yeah, it sort of does.
On the other hand, no cats were annoyed
in the making of this photo.
(18:29):
Now, I want everybody listening, you know,
which might be only a couple of people anyway,
to realize that we're not against technology at all.
In fact, let me flip that around and say,
I am a big user of AI now for my software development.
You know, and I have to say, it's been, you know,
(18:52):
and I hate to use the phrase life changer
because, you know, a lot of technology
gets thrown around that way.
But I have been very grateful for AI in my field.
And I'll tell you some examples, you know.
A lot of times in my field,
when you're doing software development,
(19:15):
you can get easily bored by repetitive things.
Like for example, you know, you have 20 variables to name,
and they are, you know, left one, right one,
you know, left two, right two.
And now the AI actually goes and suggests,
(19:38):
oh, do you want a right three, left, you know,
left three, right three, and a four and a five,
and automatically fills in the code for you.
Which is nice.
Which is nice because I used to spend a lot of time
copying and pasting, you know, doing the replays,
and sometimes something gets missed and introduces an error
because you forgot to rename something
(20:00):
that was supposed to be renamed.
And that has changed the way I develop, you know.
Another thing is a lot of software developers,
you know, we didn't have everything memorized.
You know, you're being asked, hey, can you do this,
you know, this task, blah, blah, blah.
(20:20):
And you basically have to start researching,
and you go online, and you get code snippets
from other developers who have done similar things.
Sometimes even your old code snippets
because it's been so long.
Yeah, even things that you have done before.
But as a software developer,
you rarely get asked to do the same thing twice.
So you're like, hey, we need this computer
(20:41):
to talk to this computer using this protocol
that they are going to communicate through
that you have never done before.
Okay, so you go online and you find out all these things
about this protocol and all that stuff.
And now what's happened is that the AI
is integrated into Visual Studio,
(21:04):
which is the thing I used to develop in.
And instead of me having to go out to different websites
and gather the information, it's right there,
and it knows all the information
from all the different websites.
It can suggest code snippets.
It just saves a ton of time.
Saves a ton of time with things that I, you know.
(21:28):
Will only use once.
Will only use once and have to spend time.
So it's not only good for me, but it's good for my employer
because they're not having to pay for me to.
Do all that research.
Do all that research and all that, and which site,
and what can I trust, and all that stuff.
I just get suggested, hey, this is the thing.
(21:50):
And of course, I still have to know enough as a developer
to know, oh yeah, this looks like good code,
or the AI is just way off today.
It's having a bad day, suggesting crazy stuff, you know.
And sometimes you also have to know what questions to ask
because it will only tell you what you asked for,
(22:10):
but let's suppose there is a competing technology
that, you know, and I said, hey, can you write this
in this form?
But I didn't even think to ask, wait, is this the best form
to answer this particular problem?
Or can you at least tell me to begin with,
what are my choices for which technology to use?
(22:32):
And stuff like that.
And so I'm not against AI.
You know, AI is great.
It's the.
But that's the AI doing the dishes for you
so that you have time to do the art.
Right, yes, that is right.
For you to be able to do the art, the artistic aspects of it,
(22:55):
you have someone, you know, doing the tasks,
like the washing machine.
That's the AI being a washing machine,
or a dishwasher, or something like that,
that is taking the things like, you know,
we think robots should do as tools
so that we don't have to do them
(23:16):
so that we can do other things.
Right. Yes.
Anyway, so that's been our talk on AI
as it relates to audio.
And the other thing is,
I wanted to talk about actual hardware, okay?
(23:37):
And this is relating to the things we had talked about
in the other podcast.
Because I'm having that dilemma again of,
okay, we have a good microphone now.
We feel like this is a good microphone.
We have a good.
Mixer. Set up mixer,
(23:58):
and recorder, and that kind of stuff.
Even good cables.
But the part that we haven't perfected in my mind
is the preamps.
And here's the thing.
Preamps are the things that microphones can connect to
that take the signal from the microphone.
(24:19):
And the microphone, if you're doing audio recording,
whether it's a podcast, or voiceovers,
and stuff like that,
the microphone by everyone's standards
is the most important part of the chain.
If you want to improve your audio,
people tell you all the time,
spend the money on the microphone.
(24:41):
But once you've done that.
But once you have done that,
the second step is what that microphone plugs into,
which can add character to your voice.
And that's called a preamp.
It gives power to the microphone if it requires power.
Some microphones are, sorry, this is the geeky part
of the podcast.
But some microphones are dynamic microphones,
(25:04):
which do not require phantom power.
And some are condenser microphones that require power.
So the preamp does more than just add character
to the voice.
It also can power the microphone and give it enough power
to go into the rest of your hardware.
(25:26):
Speaking of power, it also affects the dynamics.
So how loud you can be is,
you normally affect that on the preamps.
You say, for instance, Jose often speaks louder than me.
And so usually on the preamps,
we don't need to boost him as much,
but we tend to boost me a little more
(25:47):
because I tend to have less volume naturally.
Right.
And so, okay, so right away, I think,
and this is the stuff that always bothers me.
It's like, I have a good microphone,
but am I getting the most out of that microphone?
I spent all this money on it.
(26:07):
It seems a shame to not take the next step
and spend just a little bit more to get that.
Don't believe him when he says little.
Anyway, so I look online and right away I'm like,
well, what, you know.
What's the best preamp?
That's what he looks for.
(26:28):
What's the best preamp?
And it seems like a lot of people say the Neve 1073
is the best preamp ever invented.
And again, hardware, audio hardware is a funny thing
because everybody seems to want the old sound
(26:49):
rather than the newer stuff.
Yeah, it's so opposite of video.
It's so opposite of video.
It's so opposite of animation
and all the other stuff we do.
Which is kind of nice because it means
once we get past this hole,
I have to have the best of everything.
It should be stable.
It should be stable.
It should last for a while.
(27:09):
I mean, you think about studios like Pixar
that are doing voiceover for their animations.
If you look at the list of hardware
and you can find that online,
it is like old, old stuff.
You know, a lot of it is old stuff.
And the stuff that's new is made
to sound exactly like the old.
Yes, you can buy plugins to make it sound
(27:32):
like the old stuff.
You know, it won't sound quite the same,
but you can. Pretty close.
But people are trying to imitate that old sound
from the 70s with the hardware.
And a lot of preamps fall into these two categories.
It can either be a 1073 imitation
and there are some other preamps that are API.
(27:55):
It can imitate those.
Anyway, so here I am again.
It's like, do I spend the money on the best preamp?
And my mind fools itself into thinking that's it.
You know, that's the last thing I could ever want
(28:16):
with this in order to do whatever.
And I keep thinking about,
this is the kid with the good shoes.
Right.
Are we at the point where the bottleneck is the equipment
or are we at the point where the bottleneck is us?
(28:38):
Let me tell you this thing,
because I think maybe the kid with the good shoes
came from my dad.
This is the way that we were always playing basketball
as little kids.
We were in basketball leagues, you know,
and I always wanted the leather, really good shoes,
(28:59):
you know, to play basketball.
And at the time, my parents could not afford that
and they would buy me the cheaper cloth shoes.
I think they were Converse, you know,
and now they're like vintage, you know,
(29:22):
and they're expensive,
but at the time they were the cheaper ones.
But the point is, yes,
that it was the more affordable shoe
as opposed to the high-end shoe.
And my dad always said,
look, you don't want to be the kid in the team
that has the most expensive shoes,
but doesn't know how to play basketball.
(29:42):
You don't want to be the kid who is like super nice shoes,
but is sitting at the bench, okay?
So why don't you get cheaper shoes?
And learn how to play basketball.
Play and learn how to play
and be the best player, you know,
with just regular shoes.
(30:05):
And that, for some reason, has always been in my mind,
like, yes, it's nice to get the really good equipment,
but somebody who is skilled at doing voiceovers
can probably stand in front of any microphone
and give a performance that's memorable
(30:28):
where people don't even mind the microphone they have.
And somebody-
As long as it's not too bad.
There is a level at which the microphone is the bottom-
There is a level of quality
where you go into YouTube and hear something
and go, oh, you know, this is just, this is just bad.
Unlistenable.
But there is a point where how you apply
(30:50):
what is called compression, equalizer, all that stuff.
That's an art in itself.
And not just the performance, but the audio engineering.
The audio engineering is an art in itself.
So us not having that skill, it's easier to go,
you know, if I get the nice shoes, I'll be a better player.
(31:11):
You know, if I get the preamp,
that is considered the best in the world.
Because you can order the preamp today,
get it next week, and that's instantly fixed.
But to learn how to do audio engineering well
takes longer than a week.
And yet we have not to repeat a lot of what we said before.
(31:36):
There are times when definitely good hardware
makes a huge difference.
Even if you don't know a lot.
Even if you don't know a lot.
And even if you do know a lot, right?
Like subtypes, like we said,
the example of us doing green screening,
that we tried everything we tried.
And at some point I feel like we were really good
(31:59):
at the skill of green screening.
Until we upgraded our camera and went,
oh, we didn't have to know half of the stuff.
Half of the problems we were having was the camera.
Not being able to have enough color information
to do proper green screening.
And when we finally upgraded the camera,
(32:22):
the green screening became so much easier.
Became so much easier.
It made it look like we were doing so much better
when really our skills had not increased that time.
We had just upgraded the camera.
And so we're at that point again.
Is it good, you know,
is it important for us to upgrade that preamp?
(32:43):
Is it at the point where people who are going
to be watching our videos on YouTube won't even know.
There is some subconscious thing to sound
where people can tell this is a high production thing
because of the way it sounds.
I mean, you listen to a Pixar movie
(33:05):
versus a YouTube animation.
And you can tell right away, oh my gosh.
And it's interesting, right, because it's a lot of that
is the knowledge.
The knowing, we were looking at a clip from Toy Story 4
where Woody is talking to Bo under the car
and it's raining outside
(33:26):
and they didn't include the sound of the rain.
And I think as an amateur,
you would always include the sound of the rain.
You would.
But they just had the voices
and then they had really low some background music
because it wasn't important that you heard the rain.
You could see the rain.
It was important that you were in the moment with the voices
(33:48):
and knowing when to edit it.
And then later with, he's talking to Sporky.
Yeah, Forky. Forky.
And they added just a touch of the crickets in the background
because it was a night scene.
And then the music came up and sort of drowned that out
but the music was still really low
and the focus was still on the voices.
(34:09):
And knowing when to add those sounds.
Yeah, and that's probably another topic
for another podcast.
The idea that there are some things
that are technically correct
and you do different things for the art.
(34:29):
And that's, you know,
and then we're getting past sound engineering
and into sound design.
Yes.
And it's the evoking emotion, right?
The same thing bringing it back to AI.
Right.
AI might've put that rain in because it's raining.
So it might've put that sound in.
But the creators of Toy Story wanted to evoke
(34:52):
a certain emotion in that conversation
that was, they found the rain being distracting.
And we have realized the same thing with lighting too.
Yes.
You know, oh wait, we can't have a light here
because it's not motivated by any lamp
or anything like that.
Well, you know what?
A lot of times if you watch a Pixar movie,
(35:13):
there's light where there's no way
that there could be light there.
Just about any good movie,
there's moments where they cheat the light
because they realize it's more important
for you to see the actor's performance on their faces.
To see their faces, to see the emotion on their faces.
And so going back to Toy Story 4,
we realized the lighting, you know,
(35:35):
the rain is being lit in a way
that you would have to have lights coming from underneath.
Right.
Where there is no way that could be coming
from the street or anything like that.
Right.
And yet it's there, it looks beautiful.
But if you start to think about all the times
(35:56):
that the faces are lit in a particular way,
that there's no way that there, in the physical world,
there would be such a thing.
But you take those, you know,
and it's not a crazy thing where it's a spotlight,
you know, whatever.
Yeah, I mean, it has to be done smartly and subtly.
Subtly, yes.
Not bringing you out of the story.
But yet it's that thing where you're saying,
(36:18):
I'm trying to evoke this emotion
and I'm gonna use all my tools.
And it does not have to be technically perfect.
Right.
Like there may be a lamp
and your brain is like, oh yeah,
that light's coming from the lamp.
But when you think about it,
the physics of the way light moves,
there's no way the light from the lamp
would be lighting up their face that way.
But it's close enough that your brain is like...
(36:43):
Relaxes about it.
Relaxes about it.
There's light, I can see their face, there's a lamp.
And now I can see their expressions
and now I can feel their emotion.
Yeah.
And that's a lot better than being technically correct.
Anyway, we're gonna wrap this up
because the kids are gonna come anytime now.
And it's time to get to work.
And it's time to get to work.
Well, thanks for tuning in.