Tim: [00:00:00] I'm Tim Panton
Vim: [00:00:01] and I'm Vimla Appadoo
Tim: [00:00:02] and this is the distributed future podcast where we interview people who you might not have heard from to gain a sort of insight into what the future of technology and Society might look like. So this episode is about voice.
It's about not necessarily owning your own voice. So, you know voice is very personal but we see more and more kind of machines that talk and but have embedded human or near human voices. So I thought it'd be interesting to kind of do some questioning around that space
Vim: [00:00:35] for sure. Did I did you talk about the kind of Ethics around it?
So the we've talked a lot about. The data and privacy on the podcast and does voice fall into that at all.
Tim: [00:00:48] Well, yeah, one of the things so excited. I talked to the interviews with Allison Smith who does started out doing voice recordings for like telephone exchanges and or you know answering machines and that sort of thing and she sort of slowly drifted into more and more intelligent machines and less and less strongly scripted in some senses.
And so one of the. That we were talking about is like, you know when something sounds like her it isn't necessarily her because one of the things that she's had done is she got paid to have her voice recorded in and it's now possible to type something and I'll probably run put a demo together into this podcast.
You can type something and it will sound quite like Alison. So how do you know that that's her and you know, she has a bunch of quite funny anecdotes about like machines with her voice telling her to do things which is kind of weird, but the other thing is so it's about provenance like, you know, where did that sound come from and was it like did people have permission to use it did people know it was being recorded?
Vim: [00:01:56] And yeah, and and it's where she's letting people type in her voice. Does she care that her voice might be used for? Possibly quite dangerous things.
Tim: [00:02:07] Well her view is that it's it's so we talked about that not about danger but about like things that wouldn't necessarily be ideal like she talks about Banking and that kind of stuff authentication type stuff, but but her view it's still really entirely possible to tell that it's not her like if you know what she sounds like and I this is true because I've met her and obviously talk to her real life and it.
Is possible to tell that this is this is older technology, but you can tell that it's not actually her once it's said a few words. Hmm, but I'm not sure that in the future that's going to continue being true.
Vim: [00:02:47] Yeah. Yeah, exactly. And that what made her do it in the first place. I feel like I'm interviewing you about the interview now, right?
Tim: [00:02:53] Right. Right, right. Well, so I mean the history of it's quite and we don't really talk too much about the history because we tend not to in this podcast, but just like for context. Did she was doing like voice recordings for telephone exchanges and you know, like "you're the first fourth person in the queue" or you know, "that line is engaged" or whatever and and she got offered some work to record them for an open-source telephone system.
Vim: [00:03:22] Yeah,
Tim: [00:03:22] and that meant their condition was like they pay her to do the voice recordings, but the results had to be open source. So anyone could use these prompts so there was then this this. Really useful kind of pile of high-quality recordings of a professional voice actor saying, you know, one two three hundred thousand and you know, this line is engaged in a lot of things that you would quite nice want to use to construct this so she's got experience of like those whole words being put out there and being used and open sourced.
And so when somebody came to her and said, we're like we'd like to kind of embed you into a program that we can type into she'd already got the experience that actually the open source thing a generated her a lot of work. And so she was kind of comfortable that commercially it was going to work which is interesting because like first idea one might have is that it wouldn't but she was comfortable that he did and what she would we talk about that as well.
Vim: [00:04:21] So I when I was at a conference on Tuesday cool disruption Summit Europe and one of the the founder of what three words. Was that so the mapping technology?
Tim: [00:04:32] Oh, yeah,
Vim: [00:04:33] the world is mapped to a meter Square any every square meter has a combination of three random words associated with it. And the aim is that you can use those three words to tell an ambulance what's happening or put it into a uber, for example, and what's really interesting is they've just partnered with Mercedes to to do sat-nav navigation with
what three words.
So you would say
"Hello Mercedes what three words car house lamp"
and it will take you to the location of for that
specific thing because I didn't realize how difficult sat-nav technology is to do voice control with.
Tim: [00:05:14] Yeah. Yeah. Well a lot of stuff in the car is actually quite getting quite fiddly to kind of get it to do the right thing.
I mean, what's three what I've met them? I think it probably the met the same. One of the other Founders had in there, but by met the what three words people as well and I there's some problem with them making a proprietary naming scheme. It's kind of almost like your house number is you can only use your house number if you pay the council or the mapping company the right rights to it.
So it's a little bit like when I get it and I get. I get that they need to make money out of it and I get that it needs to be done like organized and these things cost money and whatever but it's sort of a bit like I mean I suppose in a way it's like the Ordnance survey where the maps you have to pay for them, but it does so some uneasiness in the in the open source mapping Community about the idea that this is now the net than address of your house is in some way proprietary.
Vim: [00:06:24] Ya know I've had. Similar things around that as well and the kind of the using it in the public sector and it not being completely open source and a lot of things like that around it and actually just the reality of how big is the actual problem. Is it is it that difficult to tell how far away are we from accurate voice detection?
Tim: [00:06:46] Yeah. I mean depends on what your accent like and how as another thing we crops up in the. Session is it like it's all very well saying. Yeah string these things are you know recognize you but do they only recognize you if you speak with like a classic English accent or or you know sound like the queen or whatever, you know, and I think that's an eight.
You're starting to see children who are and also the sentence construction is weird . I was at IFA the electronic show this week and I noticed. There's a coffee machine where if you wanted it to give you coffee, you had to say to it "Alexa ask the Miele the make me an espresso" and like I'm really feeling like that's like I have to name check two brands in order to get my coffee machine to work.
Vim: [00:07:41] Yeah know, not just two three if it's an nespresso.
Tim: [00:07:45] Yeah, I wish thinking generically, but there you go. See.
Vim: [00:07:50] Yeah, that's really interesting. Yeah. I wonder if it will change the. We use language, which I
Tim: [00:07:57] I fear it will . Yeah. Well,
Vim: [00:08:00] it's interesting because I think you can start to see children picking up on how their parents talk to Alexa and how you are meant to communicate to people
Tim: [00:08:10] right do you shout at them do you boss them around. Do you say please ?
Vim: [00:08:15] Yeah, is this how you get what you want?
Tim: [00:08:18] Yeah. It's a lovely thing which I'm not sure I totally believe but the lovely thing about apparently parrots will now say if they're if they're not getting enough attention, they'll say "Alexa stop".
Vim: [00:08:30] Oh wow, that's interesting.
Tim: [00:08:34] Right and they know that that'll get them get them full attention back funny.
Vim: [00:08:42] Yeah, I do think the voice is definitely becoming more a part of our everyday life and it's only going to proliferate
Tim: [00:08:50] so it's kind of weird because our work lives that got more text based and less talking based, you know, and and now suddenly Voice is coming back but not really at work.
Like very few of these systems are designed for work there for like in the kitchen or in the car.
Vim: [00:09:06] Yeah. It's almost like a how to make your home more productive.
Tim: [00:09:10] Yeah. I don't really want a productive home. I want to comfortable homes, but like maybe I'm just getting old
Vim: [00:09:16] so I think yeah one hand. You've got it trying to make your home more productive. But then on the other hand, it's how can I make life as easy as possible for you? So speak to me to get your shopping speak to me to turn your lights on speak to me to set the heating etcetera Etc.
Tim: [00:09:30] Yeah, I mean I think some of those things make sense. I mean, I think this is sort of but. Your home automation, which works for me? I mean like, you know, I knew if you're carrying shopping. It's great for the lights to come on without you having a kind of put it down and do the light switch or whatever.
Yeah, so I think they're kind of specific things where it's really useful. I guess the issue is that they're different for everybody like the different for different people.
Vim: [00:09:54] Yeah.
I'm really excited to listen to the podcast.
Tim: [00:09:57] Yeah, and I think I think the voice is the future but whether it's going to be automated voices or not.
We'll see okeydoke. We'll let you listen to it then.
Allison: [00:10:08] Hi there. I'm Allison Smith and I am a professional telephone AI voice.
Tim: [00:10:14] So an AI voice this is very trendy these days that all of the devices now have voices. So you you seeing a burgeoning market voices for devices.
Allison: [00:10:26] Yeah. I mean my core competency is and probably always will be ivr because I don't think it's going anywhere but I am making that transition into voicing activated home based platforms such as Amazon Alexa, so my voice is being developed as an upgrade for Amazon Alexa to do consumer surveys. So the idea is you shop at Best Buy and you know Costco places like that and the cashier always says if you go to our website and take our survey you can enter to win a draw and you're like no I'm probably not going to do that.
But I think people will be more inclined once they get home and they're taking their shoes off to trigger the Alexa voice and do a brief survey, you know five questions in your done and that will be my voice. Hopefully the upgraded voice so very exciting.
Tim: [00:11:19] So is it weird to have your voice coming out of the inanimate objects?
I mean you must be comfortable with it by now.
Allison: [00:11:27] I'm comfortable and somewhat detached from my voice because all I do all day every day is edit my sound files. So it's almost like a product that I produce and I'm pretty immune to the sound of it. But yes, it is extremely strange. Um to encounter my voice I always tell the story of ordering a hotel wake-up call in a hotel and it was me doing the automated wake up service.
That was
Tim: [00:11:54] Were you very bossy with yourself? Very strict must get up
Allison: [00:11:59] No, I was like Alison it's time to get up and explore the city. No it was very very soothing but as I was surfacing I was going wait a minute that's me, that's extremely strange. What do you what's weird
Tim: [00:12:12] you have do it then? There's no choice really?
Allison: [00:12:14] Yeah, what's weird is I was attempting to order flowers for somebody in the US. I was phoning on some sort of floral net work that used my voice as a, you know, a speech platform and I very clearly intoned the name of the city that I wanted to send it to. You know, I said, I believe it was New Orleans.
So I said extremely clearly New Orleans when I prompted me what city I wanted to send the flowers too. And then the automated me said okay, I think you said Madison, Wisconsin and it's like no! How are you not understanding me. I'm you
you're me
Tim: [00:12:53] the very least you could do is understand me.
Allison: [00:12:56] Yeah, that would be a good thing.
Tim: [00:12:59] Yeah, so it's kind of weird this whole business of voice. I occasionally get this thing where and I'm sort of nothing in like in your league, but mildly Niche famous in very small necks of the woods and occasionally. I come across people who recognize me from a YouTube video they've watched .
Allison: [00:13:20] Well, you're famous.
Tim: [00:13:22] Well only slightly, but it's very disconcerting because they think they know me. And I'm now getting to the point where I'm old enough that I'm not sure whether I've met them or not. So we have this kind of gentle dance around like who's met who and then I finally realized that actually they've only ever seen me on or even the heard me on it on a YouTube so or on yeah a podcast or something,
which is
Allison: [00:13:48] odd.

I've noticed at parties if I tell people what I do they go. Oh, yeah, actually, you know what? I'm pretty sure I've heard you but if I hadn't told them what I do, they would probably just think oh, that's a lady with maybe a pleasant voice. But once I tell them what I do they go I'm convinced I've heard you on systems.
Tim: [00:14:04] Right? Right. I suppose that's there's a sort of there's a comforting in the fact that you're I mean, it's I suppose it's the art really that you've got the voice to the point where it's not unexpected. I mean if people remembered it because it was kind of Brian Blessed something you'd be distracting from the from the product or the task in hand?
Allison: [00:14:22] Sure.
Yeah. Oh, yeah, absolutely. And the downside is that some people have an extremely strong aversion to recorded voices. So occasionally when I tell people what I do the reaction is I hate those I hate those automated voices on telephone. So I usually just try to fake a nosebleed and get out of there.
So but I think the reason why some people have a strong reaction is that the the automated Essence on the phone system is so detached and it seems like it's almost at odds with the caller, you know, the caller is almost trying to convince the automated voice that they need to get through to a live person.
So it's almost as though the voice is a bit of an obstacle. So I try to make the voices personable and not a challenge to callers. Is and not a you know, a Velvet Rope that keeps them back from customer care.
Tim: [00:15:13] Hello economically often that's the case sure that you know, well, I'm in this place from everybody's point of view.
If your inquiry of shipping inquiry or whatever it is can be dealt with by the ivr then then that's going to be much the most efficient thing for everyone I guess so it's a win if it works but it's kind of it's those falling between two camels or whatever it is that that you know, where the neither of the options that I've been offered like quite fit the thing that I want to do.
No, I neither want to close my account nor open a new one.
Allison: [00:15:51] And
if you choose the wrong thing you're going to be sent back through the Purgatory of the ivr to go through the whole thing again.
Tim: [00:15:58] Yeah, you bet. That's the fear always and whether it's actually true or not. You don't know you but these kind of you feel like you've invested in getting here as well.
So yeah,
Allison: [00:16:06] absolutely.
Tim: [00:16:07] Yeah, how, you you talked a little bit about, you know being in in other platforms and than than just telephony I remember that you did thing in telephony where you had your voice sampled and then people could like type a message and Alison would read it did that how did that come about and how did that sort of work?
Allison: [00:16:30] You know, it was a company called Cepstral that attended astricon really early on in Astricon. It could have been like 05 or 06 and they approached me about doing pretty much like a synthesized version of myself and I do remember the asterisk guys saying oh, you know. Into actually voice yourself out of a job if you do this, and you know, so I went ahead and I voice this gigantic compendium of you know is basically just sentences that really didn't make all that much sense.
It would be he "selected the blue Highlight marker because it was Wednesday" or something that doesn't really make all that much sense. But from all these weird sentences, they feed it into a strange machine and it separates it into phonemes and sub phonemes and hopefully they have captured all the various humans sounds or most of them.
So yeah, so it's actually their number one seller the Allison voice and it's pretty good, but it will never replace anything that's recorded from the ground up. I had a client that actually put together an entire on hold system. It was like five to seven paragraphs of spoken word with music mix behind it.
And it was all done with the Cepstral Allison voices sounded extremely choppy and very very fake I guess would be the word. So yeah.
No, it's not going put me out of business.
Tim: [00:17:54] How many words do you have to hear before you know that it's the fake Allison rather than a recording you've done.
Allison: [00:18:01] Yeah, that's that's a great question.
I think it's actually pretty early on. There's something called Prosody or Prasad e and it's the way the rhythm of how we speak and we don't make a decision to do anything like that. It just happens naturally in human speech and if it's even a little bit off the human ear can tell pretty pretty quickly one of my favorite
systems that's actually put together with text-to-speech are the airport warnings that you hear is usually an alternating male and female voice asking you to keep an eye on your luggage and stuff. Like that's usually Homeland Security messages in US airports. They're fantastic. They're actually really really well done.
But there is something that's just not quite right. There's something about the Rhythm that they're speaking that still sounds not human
if that makes sense.
Tim: [00:18:52] Yeah. I mean, I know that in some cases that. That's a deliberate decision. And I think I remember hearing an interview with the person who's responsible for voice at Google and he was saying that and he was contrasting it with Siri the Google assistants voice with the Siri voice and he was saying that they deliberately made it more recognizably a computer voice because they wanted to come over as not your friend but a giver of facts what so it was a different, you know, it's kind of almost consciously.
making it sound not like unintelligible or anything, but but not totally human. So you knew immediately that this was a computer-generated
Allison: [00:19:34] right?
Well, you know and it's interesting how far it's come if you actually if you just Google Eliza voice you can actually play with one of the original Eliza interfaces and it's extremely robotic.
It's very very fake and then they've sort of blur the line between fake and a human and. Yeah, that's that's always the debate is do you want this thing to be so friendly that they sound like your best friend? I don't think so. I think there's a real danger in it being almost to humans sounding right
Tim: [00:20:07] right.
I mean it if you remember it, but back in the long time back in the day there was a thing
called the dectalker, which
is one of the original text to speech things and somebody did this absolutely brilliant thing with it and they made it scat singing. Absolutely. You know, they just had it saying dude.
Do do whatever but but they don't put it through a vocoder and they had it's scat singing and it's absolutely wonderful dig up a link and put it in the show notes because it was glorious that so yeah. Yeah tracking back to the thing about about your voice samples and it potentially but in the end not impacting your real business the rest of your business.
It was a funny thing for the open source people to say because like it's kind of exactly their model that you throw it out there and then more people come back for consultancy. It's sort of like, how did they not see it as like intrinsically the way that they were already working or I mean slightly puzzled by that response.
Allison: [00:21:04] Yeah. Well, you know when I was first brought on to voice. What turned into asterisk it was like a really long time ago. I had a very rudimentary home studio and I was approached by a couple of Digium staffers who are long gone not dead, but just not there anymore and they were fun to interact with and the prompts were really hilarious.
You know, like some of them were serious, but some were just kind of crazy and I just viewed it as a one-off. I just thought oh, well, what fun this is so great. I can tell people I worked with clients all the way over in Alabama. That's so weird. And I literally did not think about the scope of the project nor was I told anything about how big it would grow into and.
I was invited to the first Astricon and 2004 and I just remember giving myself a pep talk going. Okay, you're going to have to introduce yourself to people because they're not going to know who you are. And you know, I struggle with shyness. Well, you know, I was pretty much mobbed when I showed up.
It was crazy. I wasn't prepared for that type of reception. It was extremely strange.
Tim: [00:22:09] Well, I mean what what you've got to understand is everyone associates hearing your voice with having successfully got this. Thing built and back in those days. It was it was like, you know, you could spend you could spend an hour or two or a day or two depending on how weird the hardware was getting it working.
And then when it finally worked you dial, what was it six hundred and they would be Allison saying congratulations, you've successfully installed and it was like it's his victory moment. So everyone in that room associating you with success.
Allison: [00:22:40] Congratulations, you have successfully installed the asterisk open source PBX.
Tim: [00:22:46] that's what they heard. Yeah. I know I still think that weasels have eaten our phone system is my favorite. Yep, but that I've got a was got a telephone new set of telephone lines just many years ago. New set telephone lines into the office and and they delivered. There was something weird about how they delivered the the connection and so I plugged in this asterisk box and left it and then I got a phone call a couple of days later saying when we think it's working with us something very weird about it.
And what I forgotten is I'd left the test messages being weasels have eating our phone system. So this very concerned man from BT rang me up and like. Well, what happened to me? So yeah,
Allison: [00:23:29] you know the other story from the first Astricon is there was one of the weird prompts that I recorded. That was I hope I can say this on this podcast moose penis.
I don't know why but they wanted me to record moose penis, but I had forgotten that I recorded it. So at Astricon the very first one a group of guys came towards me and they were all yelling in unison moose penis and I thought well, this must be a strange Atlanta greeting. I am not sure what this is.
Yeah, it's
Tim: [00:23:57] I can't think of a justification for that but never mind it. So yeah, I don't even know the reason that some point
Allison: [00:24:04] where in the sequence would that show up? I don't know.
Tim: [00:24:08] What number would you dial for it?
Allison: [00:24:10] Yeah, exactly. But you know that whole Community has just been incredible. And as you say me doing the open-source stock prompts obviously led to a great deal of recurring business for people who needed customized recordings and it still does it's died down a lot.
People are not using asterisk as an ivr so much. So yeah, the call for a customized prompts is less than it was. You know, let's say 10 years ago, but it's still, you know a thing and I still interact a lot. In fact today. I have to record a little pre-recorded video greeting for a group of Russian asterisk enthusiasts who are having their little gathering so, you know, it's cool.
It's still very cool.
Tim: [00:24:53] And I mean,
it's interesting that like, yeah. Source mindset and let's call it business model is actually now spreading out from or had spread out from just software to and you know in your case voice files, but you know, the whole Creative Commons movement is is a is an attempt to you know, take artistic Endeavor and codify what it is that you can and can't do with it freely and I think that's really interesting and then like I was the other day I was at an open-source Hardware conference where people are saying like, you know, the.
Designs for this thing are open source, you can you know, you can go away and build it exactly to this specification. And that's just fine which is right, you know that whole area is expanding and expanding into more and more. I want to say niches almost but like they're not niches their kind of I'm waiting for the first open source fridge or something.
I have actually seen first actually seen the first or not. Maybe not even the first place seen my first open source microscope.
Allison: [00:25:56] Really?
Tim: [00:25:56] It's a design. It's basically kind of bunch of cardboard and glass but you can build this microscope. That was know. What was it for it was for identifying malaria sufferers.
So it's like it's a cheap microscope that you couldn't you drop a smartphone on one end of it and it analyzes whether there's malaria in the blood.
Allison: [00:26:17] Thats really neat
Tim: [00:26:18] it isn't it isn't it? But just like that whole open source ethos spreading out from where it began and out into you know all.
Lots of places which is kind of fun.
Allison: [00:26:28] Well, I'm just the benefit of the idea of bug tracking an entire Community sharing solutions to common problems is a huge huge thing, you know, if anyone were to encounter bugs or problems within an open source system, there's obviously huge resources. They're already working on the problem.
So ya know it's amazing.
Tim: [00:26:50] Yeah. It's a funny funny space. I mean, I've really enjoyed it watching it grow and happen around us. I was gonna say it was just backtracking a long way you were talking about prosody whether I pronounce that correctly or not, and I wanted to kind of I don't know if you do like listen to many podcasts, but there's a really nice one called lingthusiasm, which I was listening to the other day and they're interested in in language, but they were talking specifically in an episode of couple months ago about gestures and there's this really amazing thing which.
That if you're going to say, oh he's over there then it turns out that your hands start moving to the so that you will point at the right moment in the sentence so in order to make the gesture of like the ball is rolling down the hill or whatever. You've already moved your hands to the right place to do that before you've got to saying the words before you probably even thought about what that word is obviously not totally but like your hands are correctly placed to do that, which is really strange.
So there's a kind of huge bunch of processing going on there.
Allison: [00:27:55] And you're talking to somebody who regularly knocks over her microphone because I do gesture as I'm recording. I can't help it. I'm gesturing right now as I'm talking to you. So and I'm not even Italian so ya know it's I'm going to look that up lingthusiasm. . Yeah, it sounds great.
Tim: [00:28:11] It's cool right there and they love it. The thing that you mean the thing I like about them is that they're absolutely kind of obsessed with it with language in the formation of language and you know other peoples languages and whatever it's. So
Allison: [00:28:24] yeah, and and the cultural directions that language take I kind of have an interesting sideline and that's vocal coaching.
So I'm trying to coach particularly young women to stop doing things like up speak where the ends of sentences. Sound like a question. Maybe that's a North American thing. I don't know if the air is
Tim: [00:28:43] not a really straight and the Australians to it the whole time. I did. Yes. No, it's kind of I mean, I'm no expert but it's a I think that I almost feel like that's where it came from but I don't know.
Allison: [00:28:55] Well,
there's some discussion about it being a Canadian thing too. Because you know, we're an apologetic polite people and perhaps we have this tendency to maybe talk in sentences because we're afraid to make declarative. The you know declarative ends to our sentences. So
Tim: [00:29:12] hmm. Okay, so it's kind of takes the sting out of the assertion or something.
Allison: [00:29:16] It does. Yeah, and you know, I started coaching yoga instructors because I'm a big yoga practitioner and yoga teachers do not want to come across as bossy or you know saying do this now put your left foot forward. They don't want to come across as being strident. So they they temper what they're saying with a lot of delegates questions sounding things that aren't even.
Tim: [00:29:38] But they still want like I mean, the whole point of presumably of going to the class is to be taught things. So it's not the know. Yeah, it's not that you're not there for it. If you see what I mean,
Allison: [00:29:49] exactly exactly.
Tim: [00:29:51] So, where do you see kind of spoken language going? Is it something that you think is like kind of I mean, I've spent the day in a co-working space where pretty much everyone was typing there were a few people on the phone, but really very few.
It was almost typing. Do you think voice is kind of Disappearing completely ?
Allison: [00:30:11] Not completely but I do see a trend towards less typing especially on smartphones. It's I find that I myself if there's a microphone option to dictate it into the system as opposed to typing it because I'm a terrible typist.
I will go with voice activation. But yeah, I think that's probably the trend is to go more towards a spoken word shortcut. It's just less labor-intensive. As we become lazier, I think it'll just be I mean my three-year-old great-nephew already knows that if he wants to watch television, he first has to say OK Google and activate the TV.
Oh and I just activated my Google on my phone
Tim: [00:30:57] yeah. Well, that's the problem. You need that. They're not they don't know whether it's them being spoken about or not. Like you need to you need to somehow refer to them in the. At present to escape that that issue you can otherwise,
Allison: [00:31:10] know always the question always will be will the older population be left behind.
I recently just voiced an ivr and they wanted each recording to end with the phrase. This is a recording which is kind of strange to me now in 2019. Do we need to say this is a recording but you know, it's possible that the elderly will say. Well I just talked to a gal who said. Who's going to look after this issue?
No, you weren't talking to a gal. You're talking to a voice of a gal. So yeah, it's interesting to find out if there's a certain set of the population that's going to be left behind with the technology.
Tim: [00:31:50] There's a really terrifying thing about these these ivrs that are constructed to kind of, you know sell you into something or whatever where the the ones that actually have a kind of an AI behind them those.
Apparently if you ask them with natural language processing if you ask them, "are you a machine?" They laugh!
Allison: [00:32:12] Yeah, which
Tim: [00:32:14] is like, you know, because they don't want to actually lie because that would invalidate the contract with the only other hand. They don't want to let on that they aren't human beings because that's not the kind of role they're playing.
And conversely like, you know, if a human being who's tied to a really strict script isn't functionally different really because they've gotten out of the options. They've got no greater.
Allison: [00:32:38] Absolutely. Yeah, huh. That's interesting.
Tim: [00:32:41] It's depressing actually in an odd sort of way, but you know, but I mean, I think I think about your your issue of saying well, I'm a recording I think that partly.
I mean, I know if you've noticed this but on the phone in particular the audio quality is much worse than it was kind of went through a peak landlines. I think went through a peak about 10 years ago quality and from my perception of phone quality is now going down.
Allison: [00:33:04] Yes. Yeah
Tim: [00:33:05] and that you you know, if you're making traditional phone call to an ivr the quality may actually not be good enough for you to be able to tell whether it's a Well recorded Voice or a human who's being strongly script.
Allison: [00:33:16] Right exactly. You know when I record I always recorded really high resolution files. But then I also I take those files and I downsample them into a telephony format that they can use but I was remember years ago. One of the digium staffers said record in 48,000 hurts instead of 44100 because files down sample easier indivisible as of 8 did you know that I didn't know that.
Tim: [00:33:44] Yeah well. To put it on put them onto a CD in which case that's no longer true. The 44k1 is so it's like where is it going? I don't know where it's going. Well, it might yeah bit of a nightmare all of that.
Allison: [00:33:59] I did have a moment where I said does 48 thousand Hertz make me sound fat?
I'm kind of worried about that.
Tim: [00:34:05] Hopeful fat is full
Allison: [00:34:08] exactly.
Tim: [00:34:09] Wide audio you know, it's not not what we're out there. We want thin audio. No, that's not right. That's not right. But I mean I do get I'm you know doing this podcast and having a bunch of people who from the kind of asterisk Community who are really into audio has been a real eye-opener.
For me, they've been like super critical of you know, where there's way too much echo in that room or you know, you should be sitting closer to the microphone or you know, these kinds of things and and honestly, I sort of when I think about how people are going to listen to this podcast they're going to listen to it while they're frying onions or or driving a car or you know, whatever and provided its intelligible and not annoying.
I don't ya kind of feel that we need to go to the full studio quality, but. But I mean, I guess that's what you spend your life doing producing a like as good at recording as you can
Allison: [00:35:04] yes, but yeah, I don't belabor it now don't overthink it too much and I actually try not to change gear all that often.
Sometimes you have to update things like, you know, audio processors and your computer and microphones. I've been known to kind of swap around but I'm very very much a creature of habit and with doing ivr are almost exclusively. Audio settings do not change from day to day. I keep everything exactly the way it is because consistency is really an important thing.
Now, of course I'm aging and I think actually my voice sounds better now than I did 20 years ago and there are some asterisk wonks that say, oh wow, you sound different now and you know, I think it's for the better but it's inevitable,
Tim: [00:35:50] you know, that's kind of odd. It's like coming, you know a set of in size, but it's no different for photographs, but but.
Many people have a like a voice record that as precise as yours that goes back as far as yours does with us higher quality as yours, that's kind of interesting them. There's a there's a kind of future history exhibiting that somewhere
there may very well be yeah, but I mean even singers I mean take a look at how Joni Mitchell sounds now as opposed to in the 60s.
She has such a gravitas in her voice now, which is a good thing. Yeah.
I went and saw a who was it? I'm trying to think. Oh, I can't remember now a Canadian singer who was talking about how doing Joni Mitchell cover versions is almost impossible because Joni Mitchell wrote for her voice and and and anybody else who tries to do it straight covers doomed pretty much.
Allison: [00:36:50] Yeah. Yeah, you have to yeah, you have to tread very carefully with people like Joni Mitchell any of the the. You know high-profile singers who have a signature sound. Yeah, I wouldn't touch covering a Joni Mitchell. I wonder who that was. Was it Sarah Mclaughlin or who wasn't it said that quote
Tim: [00:37:08] gone.
It'll it'll come back to me in in 10 minutes time. But you know, that's that's the risk of bringing up something random without having looked like yeah.
Allison: [00:37:20] Yeah, so I found different at different times of the year to Tim. I find in the winter, especially, you know here in Calgary. Where it's cold and it's dry.
I struggle a lot with sort of a thinner weedier sounding voice in the winter. And I sound better in the summer and the spring and yeah,
Tim: [00:37:38] is that is that to do with the air or is that your mood?
Allison: [00:37:42] No, I think it's dryness Alberta has extremely dry. And so we have humidifiers going all the time and it helps but
Tim: [00:37:52] yeah, that's interesting.
I hadn't no well let those there's another thing I should. Like add to my repertoire when I build the podcast Studio of the future it should have a humidifier in it then
Allison: [00:38:04] a quiet one though.
Tim: [00:38:05] Well, yeah, that's the problem. Right? How do you yeah, yes. Yes indeed. No, I think I think there's a actually the thing that I had the other thing I've learned on this podcast, which is it.
I sort of knew it but what was fascinating is the extent to which the microphones in things like iPads and and just generally smartphones. How much better they've got in the last 10 years
Allison: [00:38:28] huge. Yeah. Oh, yeah.
Tim: [00:38:30] Just amazing how good you know if you plunk an iPad on the table, it's not quite as good but it's almost as good as putting you know, but the range polycom conference phone on the table.
And and that's amazing. Oh, yes a real. Yeah real win that.
Allison: [00:38:46] and you know, even the Optics, you know recording video with your phone is not you know, it's as good as if. At a professional freestanding camera. It's amazing the technology. But yeah, the sound on telephone systems is always going to have challenges simply because it just gets crunched down into you know, 16-bit 8000 Hertz.
Anyhow, so this idea of making it better and better not so sure if it's worth the effort.
Tim: [00:39:16] Well since so that's the thing that wish I'm seeing a bunch of movement away from telephony and into the other platforms, you know, the Skype switch actually to be fair Skype has got a lot worse recently. But, you know the other other platforms and in those cases you can get like, you know, I did some work for Wire a while ago and the in those cases you get this fantastically good audio quality and it does make from I mean this call is is you know decent quality and.
Makes for a much easier conversation kind of the whole thing is just like less cognitive load. I find the dealing with like kind of a scratchy GSM call.
Allison: [00:39:57] Absolutely. No question. Yeah.
Tim: [00:39:59] So what do you where do you see kind of your to expect to be in the the roomba of the future or is that not a niche?
You're pursuing?
Allison: [00:40:09] No, it's very much a niche. I'm pursuing because I you know, I keep hearing talk about how the robots are taking over and. It's interesting that the Panic that people have about Automation and AI I'm just signing. I'm sort of a deciding to join the movement instead of fighting it so with me hopefully being integrated more and more into things like Alexa.
Tim: [00:40:31] You are welcoming our robot overlords.
Allison: [00:40:35] I certainly am I welcome them enthusiastically with open arms. Yeah. I
don't know
Tim: [00:40:42] excellent, glad to hear it.
Allison: [00:40:43] Yeah. Yeah. I'm going to be part of the movement and not try to fight it and yeah, it's going to happen. Anyhow Tim. So what
what can we do?
Tim: [00:40:51] I know it I'm gonna like you say it embrace it and maybe slightly shape it and I think that's the kind of you know, that's that's my thing is I kind of I want.
Certain things to not happen and I want certain things to happen. Like I don't necessarily see any reason why you know, all of the audio that spoken in my house should arrive in a server in Utah so that people can pour over it just because at some point I might say, you know, open the pod bay doors.
No, that's the wrong one, you know and make me a cup of coffee. I saw. this amazing thing. I was at the IFA the other day which is a like kind of. Conference for like home electronics, but like very future looking and and there was nothing there now, which was it it was I think it's one of the coffee maker Brands.
I want to say Mila but maybe not anyway, they had a demonstration of of Alexa. So in order to get yourself a coffee you had to say "Alexa tell the miela to make me a nespresso" and I'm thinking I have to name-check three brands to get a. Single cup of coffee
Allison: [00:42:02] right
Tim: [00:42:02] this is wrong.
Allison: [00:42:05] I'm surprised
it's that complicated.
You can't just say make me a coffee have to be so specific and especially as you say with a brand name, that's yeah, that's a little scary because they can track those metrics. There's people at Nespresso going. Oh, yeah, you know more orders coming in
Tim: [00:42:20] now. Yeah, I mean, I think that's it's sort of its like the IVR tree.
Each of those Brands is is a sort of switch point to hand the rest of the. sentence off to something else. So I guess kind of technically understand why it's there but like from a human point of view. I don't like I've only got one coffee maker. So why do I have to like name check it like if I had a variety of them spread out across the wall and maybe it would make sense.
Allison: [00:42:47] like I'm surprised you don't know.
Tim: [00:42:50] No, I'm actually I'm a reformed character and this this small flat in Berlin has meant that like acquiring random. Simple random things that I can't do clutter anymore. There's just not room for it. So I like it's me one good thing is the rule these days.
Yeah. So
Allison: [00:43:10] can I have a question for you with regards to all of these sounds that I voiced that are out there and they're going too far outlive me. What do I do about my legacy? And and how do I make sure that there's good stewardship of all of these prompts that I'm going to leave out? They're in the world.
I know it's kind of a heavy question. I'm not planning on cashing out anytime soon. Don't worry.
Tim: [00:43:35] So what I guess a lot of them already have a very clear like rights and ownership about them. So they're you know that the open source ones have well delineated what the rules are about those and I guess the ones that have gone out to Cepstral presumably Cepstral keep the rights in exactly the same way and they distribute the rights and exactly the same way whether you're involved or not. Guess right so I don't get it. I mean, I don't know if you've got other should have retained rights. If there are things that you get paid kind of like an author by by percentage or by sales or something then I guess that's that's like complicated actually
Allison: [00:44:15] it is.
Tim: [00:44:15] Yeah. Yeah. I honestly don't know. I mean there's a whole movement about talking about like what happens to your digital assets after you die and you know Facebook now have a thing where you can designate people who will be responsible for for your web for your Facebook presence after your death.
Allison: [00:44:34] Yeah,
Tim: [00:44:34] which is the thing when people don't do it. It's really disconcerting like every now I get I get a Facebook reminder that it's you know, X Y Zed birthday and I went to their funeral 18 months ago. Yeah that is like and and in some way I mean coming back to voice. I think like reading that is one thing but having it said to you would
be even worse somehow. I'm not sure why but like if you pop up in your ear saying it's you know, it's like that's like
Allison: [00:45:09] Yeah. Yeah. No, I actually had a friend who passed away and you know a couple of years after he passed there were still some of his Facebook friends. He's page was still up and running.
Some of his friends were saying hey, I haven't talked to you for a while how our thing is. I was like, oh God, they don't know and it's an awful. Yeah,
Tim: [00:45:26] it's all right. I mean that's a disconcerting piece of sort of not tied it up somehow and I guess the law hasn't taught caught up with it in a way where people's habits it.
So I don't know how that would get it done. But yeah, I mean, I guess coming back to your question. I guess most of what you've done is probably already sort of reasonably legally tied down or not, but it's like if it isn't it isn't and is never going to be I guess the one I would be kind of if I was you I would be more disconcerted about would be a future better version of cepstral.
So if you do a sort of a subsequent version of that, but that is less discernibly automated that could give like, you know, your relatives and unpleasant surprise
because it
Allison: [00:46:16] definitely could
you know Tim. I wanted to also mention that there was another text to speech. Platform that I recorded it was kind of a collaborative thing. It was a couple of guys in the asterisk community that wanted to build their own Allison voice.
So I actually went ahead and voice the Corpus they kind of got in over their heads and they weren't exactly sure how to complete it and how to. Again, take those sounds and break them into phonemes. So long story short that is available. I have it sort of archived if anyone's interested in and has the wherewithal to develop it into another text-to-speech engine voiced by me.
It's it's it's available.
Tim: [00:46:57] So actually yesterday and I mean I can think of a couple of projects where that might turn out to be useful in that all of that technology is getting much more flexible now and much much as you can always can say easier to use. I'm not sure that's true but kind of computationally some of the algorithms have got a lot smarter.
So I mean there's this thing this deep fake thing where you can give it a bunch of of your audio and then it'll make a mediocre text-to-speech automatically for you from you. It's discernibly I might do a recording and drop it in here. So that people can see what it sounds like. But but it is discernibly not the person that are the other hand.
It's sort of sounds a bit like them. See what I mean. Yeah, she's yeah, which is interesting and that's a useful useful attribute. But yeah, so yeah so we can expect to hear you in as the voice of the future coffee machine.
Allison: [00:47:50] I hope so fingers are crossed.
Tim: [00:47:53] Right excellent. Well, let's let's let's work on that.
Then let's work on that.
Allison: [00:47:58] Your coffee is ready Tim. Yeah. No I could
get into that
Tim: [00:48:02] I think it needs to be a bit sharper. Well depends
on the time of day like first thing in the morning and you should be quite kind of bright and sharp.
Allison: [00:48:10] Yeah and by 3:00 p.m. I'm very very caffeinated. Yeah. It's a different vibe all together.
Tim: [00:48:17] Yes, cool. You have to act like. Multiple recordings depending on how many cups of coffee there that it issued today. That's there's a lot of work in that
Cool. Thanks so much for your for your time on this and my apologies for the mess up that with the first one. But it's great and good to talk to you and hopefully we'll see you some point in Europe in the not too distant future.
Allison: [00:48:43] Yes,
absolutely alrighty . Thanks so much Tim.
Tim: [00:48:47] Thank you.
Allison: [00:48:48] Alright. Take care. Bye.