Voice Acting, Talent and the Future of AI Voices with David Ciccarelli

David Ciccarelli:

I think history's proven that automation actually improves workflow, creates opportunities for creativity that previously didn't exist.

Stephanie Ciccarelli:

Could your voice be making you money when you are on vacation? In this interview, you're gonna learn what AI voices are, why we even need them, like what they're used for, how we go about making them. And what if you were to have a clone of your own voice? Enjoy the show.

Stephanie Ciccarelli:

Well, thank you for joining us again on Vox Talk, David.

David Ciccarelli:

Hey, great to be here, Stephanie. Thanks for the invite back.

Stephanie Ciccarelli:

Well, there's certainly a lot of interest in AI voices, and we're gonna start at the very beginning. So, so David, what is an AI voice?

David Ciccarelli:

Well, an AI voice, uh, is probably an, an all-encompassing term. Um, some people call it synthetic voice or voice cloning. It's really the creation and replication of someone's voice synthetically through artificial intelligence from audio samples. I'm gonna say this really represents a, you know, a true techno, uh, technological advancement likely to be a meaningful opportunity, both for the members of voices.com, as well as, uh, voice actors everywhere. But there's some nuance in that, you know, how do we go about it? We wanna make sure we're doing it the right way, but it's certainly here. Uh, and, uh, we should be, uh, it's great that we're even having this conversation.

Stephanie Ciccarelli:

Right? Well, it is the talk of the town. I think we had, no one can say that it isn't, you know, turn over a rock. There's AI voice, turn over that rock. You found it again.

David:

Mm-hmm. ,

Stephanie Ciccarelli:

You know, it's obviously used in many different ways. And, and that said, you know, David, what kind of AI voices are available? Like, what are people, um, using them for and, and why?

David Ciccarelli:

Well, I think the, probably the most common, um, are these digital assistant type voices, where you think of Siri or Alexa or Google Assistant. Um, that's not exactly what I'm referring to as like, you know, kind of the, the opportunity, if you will, for talent. Um, there's, uh, but let's call those assistant voices. There's also, um, stock voices that you can go on a website, select from a library of nameless voice actors across, you know, gender, language, sometimes tone or style, but they don't, they just have an avatar. You don't really know who's behind that voice, um, so that, again, I'm, I'm a little, uh, weary of, uh, of that, uh, approach. Uh, the one that's most exciting though is I'm gonna describe it as a cloned voice, where a potential client could select, uh, from you know, professional voice actors who are known entities and hire their natural voice or work with their cloned voice. Um, perhaps, you know, time constraints, budget constraints, availability, um, but it's a copy or a clone of that actual talent. And likewise still be available across, you know, all those different, um, roles and styles that an actor can perform. The last, you know, uh, area in terms of a type of AI voice would be some large brands are actually looking to create, um, you know, a, you know, looking actually to acquire voice data to actually train and kind of create a, you know, uh, a mashup type voice. You know, again, I would be a little hesitant on, on these types of, uh, job, uh, opportunities. Only because you are not really getting credit. You're getting a payment upfront. Um, you're not getting credit on an ongoing basis for any use of your voice or every time it's used, but you're the da you know, the other thing is that you're really just providing your voice data. Um, and so know that that's kind of the arrangement that you're entering into. So of those kind of three or four I describe and actually, um, consider that the voice cloning, cuz it's really a copy of you as a talent. It's a model of you that you could work with or your clients could work with and continue to be able to generate income, uh, from people who want to hire a voice that sounds, uh, just like you.

Stephanie Ciccarelli:

Right? So a good example of say a voice clone, I'm thinking would be, um, go all the way back. Roger Ebert, um, a little more recently, Val Kilmer,

David Ciccarelli:

Mm-hmm.

Stephanie Ciccarelli:

That sort of thing, whereas the actual person. Um, but it, it seems like that's the way you could earn more money as a voice town is to essentially scale yourself, to scale your ability to record, to, to do pickups, even like this voice could do work for you that you really don't have time to, to go into the studio to do. So I'm just thinking back to when people would, you know, these big celebrity actors in studios, you know, when they go in and they do their work, they're done. And oh-oh someone made a mistake, better send in their voice match so they can go re-record. Well, this would be like the next step up. It's like, instead of even having someone go back in and re-record and sound close enough mm-hmm. , it's actually that person's voice itself. You know, voice cloning out of the options you've mentioned sounds far more advantageous to the professional voice actor than say, having your voice mixed up with a bunch of other different voices. You have no credit. Um, you get paid up front, um, you
know, the one time

David Ciccarelli:

mm-hmm.

Stephanie Ciccarelli:

and, and then your voice is in like a hodgepodge of something and, and just a little, you know, like a little, little weird, if you don't mind me saying. So,

David Ciccarelli:

Yeah, so I mean, maybe we can hone in on this concept of voice cloning. I mean, there's certainly benefits to talent. Um, the ability to offer both your natural voice and your clone voice. Uh, we talked about any time availability, um, still with the, uh, you know, an, an important attribute would actually be able to have control over the content that is produced by your clone and the brands that you work with. So, I mean, that would be a benefit to doing so versus you contribute to that mashup AI voice. You're probably not having control over the content that's getting the words that are being put into your mouth. And, and then of course, the clients, you know, who have expressed interest in this, they're, they're looking for consistency 24/7 availability. There's no kind of downtime or vacation time or, um, should someone's voice change slightly or if they're ill, you know, none of those factors kind of come into play for the clients. What we've heard is that they're looking for most is just the ability to be able to paste in a script, get a preview, and this is the important part, an instant download of the audio file. You know, the economics kind of the lower the perceived lower, uh, cost to doing so, um, was rated lower, but it's, you know, it's certainly a consideration. Um, but this concept of the instant download seemed to be of the, the greatest benefit, um, to clients. But we've run surveys, uh, on this, this whole area. Um, and, you know, a few quick stats I think are, are, uh, salient points here. You know, first off, clients are, you know, to, to underscore the point of like, well, why now? I mean, clients are already using AI voices. I mean, they're citing speed is their top reason why. When we surveyed our clients, 41% of them said that they've used an AI voice or text to speech software to create a voice over recording. That seems really high to me. So they, that they're already dabbling in this, they cited speed as the top reason why. So 44% people cited speed as the top reason why to use an AI voice versus a natural one. Those who haven't yet, I'm gonna use the word again, dabbled in, uh, working with AI voice, 60% said they would, if they could, they would consider doing so for future projects through voices.com. We almost have like 40% have already been doing this. The other 60% would do so if they could hire someone's clone through voices.com. To me, that's really indicating that there's an opportunity here and now, and that we should collectively as a community be responding to that.

Stephanie Ciccarelli:

That's really interesting. I know that in the survey you didn't necessarily know, um, what purposes they were using the AI voice for. Um, however, you didn't know that they were using it, that they've wanted to use it and they wanted it. Now it sounds like this immediate gratification sort of culture we find ourselves in. Um, but I mean, time is money and if they can save time, um, just updating that HR training video or whatever it is, they don't have to go back and say, Hey, so-and-so, and they have like, you know, a six to 24 hour gap perhaps from when they might hear back and get something. And, and then the talent would have to match what they did the last time and remember what they ate. I know there was some talent , just thinking back to an old episode, um, go back and listen. I thi I can't remember who's Anthony Reece, but somebody, somebody said maybe it was David Kaplan actually, that they actually kept a record of what they ate that day when they did a session, you know, so that they would know what went into the sound that was created. And they keep samples and journals. Pretty sure it was Kaplan. You let me know, David if I'm wrong. Um, but, you know, there's a lot going on with just, you know, the accessibility of it. It's so easy and, and it's your voice at its best at any given time, any time of day for that client to download right away and you still get paid for it.

David Ciccarelli:

Well, let's unpack kind of those use cases. You know, you're right in the, in the survey, uh, we didn't ask, you know, those how you're using AI voice, but just that they were. Um, but I would, I would argue, you know, that AI voice is most suitable for jobs that are purely informational and that really have very few words or many words in the content changes often. Okay. So think of like, let's just go on like the one end of the spectrum of emotional to informational. It's definitely on the informational, um, uh, side of things. Think of like turn by turn directions, um, phone system greetings, um, platform changes in a train station or gate changes in an airport, news weather and sports update. This is just information. No one’s looking for personality per se in, in this type of content. Um, and then, uh, it also probably has very few words and the content could change quite often.

Stephanie Ciccarelli:

Right?

David Ciccarelli:

So that's where I would say that we're already seeing that, that AI voices are entering in jobs of that, um, you know, that have those dimensions, few words in purely information. On the other end of the spectrum, Stephanie, you described kind of these big e-learning projects or corporate training videos where there's many, many words. Well, sometimes, you know, it's just not economically feasible to even do the project in the first place. You know, it's the, uh, you know, the curriculum designer who's working in the HR department, who's, you know, they have a job to add in audio, perhaps even just for accessibility purposes, but have a very limited budget to do so. And so when they come to talent and say, Hey, I've got 20,000 words and it's, you know, a few hundred bucks, most talent are gonna go, yeah, I'd rather pass. I'm absolutely not even going to take my time.
Stephanie Ciccarelli:

Run. Yeah. Right.

David Ciccarelli:

Don't even audition for it. And so those, uh, that, you know, is kind of that other end of the spectrum. Um, the many words, AI voices are similarly addressing jobs that have many words that are often just not economically feasible for voice actors.

Stephanie Ciccarelli:

Right? And, and they're not stories in a way, like, it's not storytelling. It's not, um, a character on, uh, your favorite animated feature. Um, you know, the work that the AI voices are best suited for, as you said David, are these short, communicative little bits of information that help you get through to the next thing. Like this is already happening in the visual space. It's crazy just what people are able to do with photography these days with the AI. I like, I, it's an, I don't even know what's photography. It's just art that's been made to make it look like it is, but it's happening in the sports world. I know we've talked about this in a past episode, uh, where those highlight reels that you see put together of all the players, you know, he shoots, he scores all whatever else, like you're seeing literally, like a machine knew what to do to just splice it all together. So that was, uh, the Oscar Sanchez episodes, if anyone wants to go back and find it. But, you know, there's just so much work that voice actors have that they don't have to be worried about, you know, the AI the, the robots are coming, they're taking our jobs. Like people still wanna work with real people. It's just the certain kinds of work that actually helps you. It augments your career as a voice talent to have your own voice do it, uh, for you in effect, uh, you know, and, and not to have to read everything out yourself.

David Ciccarelli:

I mean, one, one parallel might be, you know, did animation and animators take over live action actors? No. Did green screen performers and performances, you know, replace live action? No, probably augmented it, maybe added new special effects, likely accelerated workflow, and even do things that couldn't be done otherwise. So this is kind of a longstanding concern, not just in voice over, but, and not just in entertainment, but overall are, I think to use your term, kind of robots taking over the world. I think history's proven that automation actually improves workflow, creates opportunities for creativity that previously didn't exist. Uh, so there's, I, you know, that's why I'm, uh, I guess the eternal optimist on this whole theme

Stephanie Ciccarelli:

So, um, well, you know, as we go about just thinking about the space, there are a lot of players involved in voice tech, like a lot. And there are conferences. I know you speak at some of these, uh, Voices sponsors some of these in fact, and have presences there to meet people and learn and share. Um, but, you know, there are voice agencies, there are all kinds of different people out there, and, and sometimes it can be a little confusing as to know, um, who they are and what they do. So, David, could you let us know what is this ecosystem like? And, and where do we as people who provide real voices and in some cases, hopefully down the road, cloned voices, fit in?

David Ciccarelli:

Yeah, that, you're right. There's an entire ecosystem of companies that are sprouting up, uh, for that are really enabling conversational technologies. Think of, you know, chat bots to voice assistance, to, you know, that actual layer that we hear and interact with, which is that the, the, the AI voices, the, you know, variety of organizations range from, you know, what are referred to as like, you know, call it the big tech companies that all have some type of text tope, uh, offering. Then there's open source projects that are out there that are just getting started, that are doing similar thing, but make it more, you know, free and accessible and projects that other developers, uh, voice application developers can contribute to. There's companies that, you know, aggregate and package up and make available data sets for training purposes. And it could be voice data, it could be mapping data, it could be location, you know, information of, um, you know, as I, or the news weather and sports example as I, as I've given. So there's data sets that ultimately get read out, um, by the, the voices. And then there's entire, uh, language models, which sometimes are referred to as like large language models where when a word is, you know, typed out in a text to speech engine, okay, how is it supposed to sound? You know, uh, you know, what's the interpretation of that? And so there's, you know, there's definitely large language models as well. And then I would say, you know, from sonic branding, there's kind of a whole other area of audio production. So there's a lot of different companies, I would say, if, if anything, we're probably unlike this, you know, uh, you know, AI voice component where it is the piece that someone actually hears. Um, now, right now we don't have an AI voice, you know, offering in the market. It's certainly something that we are exploring. We're, uh, as you described, Stephanie, active in the space in terms of learning. Um, and for those who probably remember, I mean, we were, I think in 2017, 2018, first of, uh, in, in the industry to come out with a partnership with, uh, VocaliD, who's now, uh, part of Veritone and, uh, really early at the time, but they were one of the first ones to actually be cloning voices. Um, or initially for, you know, for medical purposes that people have lost their voice or wanted to preserve it in advance of losing their voice, but now are looking at more kind of, you know, commercial or, uh, enterprise type applications. Um, but the enablement of that, these voices don't get created outta nothing. Some, some brands, some organizations somewhere needs to actually hire a talent, um, and have them clone their voice or have them contribute their voice as data to be able to, uh, you know, create a synthetic. So I think that's, uh, where we have been playing, um, where we continue to do so we just don't have kind of, uh, at, at present, anyway, a more formalized, uh, product offering.

Stephanie Ciccarelli:

Right. And you did mention vocal, so little shout out to Dr. Rupal Patel, good work, , congratulations again on, on becoming part of the Veritone family. Uh, yeah, so there's just a lot going on in that space. We've been quite involved in it for a number of years, as you've pointed out. Um, it's an exciting time for talent, really. Um, but many are wondering, and I know they've gone from trepidation to now, curiosity, I think, you know, and kind of, uh, they're wondering how they can protect their voices and also get paid appropriately for the work. So, um, what kind of advice do you have for them?

David Ciccarelli:

Well, I would encourage talent to work with organizations that you know and trust that are going to adhere to, you know, the principles of Ethical AI. Um, I've recently joined the Council for Conversational Leadership and conversational design, uh, leadership, where we've, you know, in effect a group of almost four to 500 executives signed, you know, a letter agreeing to these principles of ethical AI resisting kind of, uh, reciting the letter, uh, and those principles here, which you can find online. Um, and we, we can link up in the show notes as well too, Stephanie, um, how we actually apply those, and we're thinking about applying those at, at Voices is, um, if the talent is considering having their voice cloned, the first thing is make sure that you're giving consent, that there's an agreement in place. You know, where a lot of artists, and I use this in terms of graphic designers or photographers, why they're upset, is because some unruly companies, dare I say, have gone out and scraped the internet and not gotten consent from the artist to be able to feed that data into the machine, so to speak. And so they've, they've done that in an, you know, arguably an unlawful manner, and then are generating, I saw just last night, um a lady on LinkedIn had generated, you know, 20 different headshots, um, of herself using one of these tools, which, you know, looked great. But that AI had to come up with those quote unquote ‘ideas’ from some other source, which was probably other professional photographers. Um, likewise with the, you know, graphic design. Um, you know, this, the kind of a generative AI artwork, those originally came, those ideas originally came from artists. So the artist, whether you're a voice actor or you're a graphic designer, or you know, content writer need to give consent to contribute to these type of programs or consent. If you are, it's going to be your voice in, in our case, actually being cloned. The next up is getting, if you've given consent, then you should also be getting credit. Hey, I'm actually the source of this voice. So that's why I like this idea of voice cloning, because it ties back to an individual, a human being, a known person where they can get credit and they could potentially even be hired for the, you know, the big full job. Maybe the AI voice or that clone, that clone, if you will, was used for the spec work, the scratch track, you know, the rough draft. But then the, you know, often in like ad agency pitches or maybe early on in the creative process within the brand marketing team, but then when they want to do the real thing, then they hire the natural voice with the actor. And you can only do that if you, if the, if that actor has credit, you know who to actually go through and hire. And then the important last part is whether it's your clone voice, um, just like with your natural one, you should be getting compensation. Now, there are those opportunities to describe where you might be contributing your voice’s data. You're at least you're giving consent and you're definitely getting compensated for that. It might be an upfront payment. You know, if a, if a company wants to use your clone voice exclusively, then it might be an annual, you know, uh, compensation and annuity where it's an annual contract that that comes up, spoken about that with a number of talent and other producers that, that, that's appealing to them. Or if it's a clone, your clone voice is on, kind of available on a per project basis. You know, it might not be in the same rates that you're used to now, but it's also, the effort is dramatically lower and there. But the key here is that, um, the talent should still be compensated every time their, uh, their clone voice is being used.

Stephanie Ciccarelli:

Right. We've had conversations, um, also with Bev Standing who we know, uh, her story, if you haven't heard it, go back and listen to that episode of Vox Talk. But, um, you know, talent are very concerned about being protected and rightfully so, and, and also choosing what their voice is going to read out, not just that their voice is able to read something out at all. Uh, so I, I think that that is really, really good that there is a statement that has been signed by four to 500 or so of you, um, you know, big companies and small companies involved in the creation of AI voice and, and ai. Um, that, yeah, everyone will definitely go back and listen to, um, the episode with Bev. Um, I won't get into the TikTok stuff, but it has to do with TikTok. But anyway, good stuff. We will link that up in the show notes. So as we are talking about the voice actors, and maybe you've piqued their interest here, they're, they're like, ‘huh, I'm actually, you know, not on the fence anymore. I am thinking this might be a good idea.’ Um, what is the process like for the talent when they do wanna go and record their voice and, and make it a clone?

David Ciccarelli:

Well, there's, uh, generally two ways to do this. Uh, one is an effect speech to speech, and the other one is text to speech. And the speech to speech is, okay, your voice, uh, preexisting voice recordings could be all of your demos, um, or could be some new content. But in effect, starting with the audio, um, that's preexisting actually loading that into, um, you know, one of these, uh, you know, machine learning systems, and it actually identifies the nuance of your voice that is different and kind of on top of this base model. But you're starting with audio, uh, pre existing audio, and this is, people might have even heard are like, oh, you can clone a voice with as little as three seconds of, of, you know, it's like, yes, because there's a base model where, um, male voices have a lot of similarities and commonalities to it, just like female voices do. And so it might, the, the base model might, uh, for cloning might get you 80, 90% of the way there. And so that's why that's this, it's so shocking that it's like you only need three to five seconds. Um, it's because it's actually only addressing the nuance and the distinct attributes of your voice. Now, generally speaking, the more data you feed it, the better quality it gets. Um, and so, you know, three seconds is an extreme, but it's recognizable. And this is where you hear, you know, um, presidents, you know, you know, doing sports commentating, uh, or, you know, reading a speech that's something that they didn't say, or, you know, playing video games, this kind of thing that you might, you may have seen online. Um, it's because they're taking clips from pre-existing audio. Obviously current and past presidents aren't going into a studio and recording anything new. They're just using pre-existing content. And the other approach is, is actually reading a script. And there are, you know, common scripts that are out there. I've actually gone through, and this might be maybe news for the community, um, we've been exploring this just as a proof of concept here at voices.com. And, and to do so, we've worked with our data sciences team, um, to figure out how, how this is done. So I've done the three to five second bit. I've also done the, you know, three hour approach, which is reading a script that exists, uh, which is a, uh, actually a government document on the JFK assassination. It's a very dry document. Stephanie, um, probably explains why a lot of the voices, you know, are often criticized of sounding very robotic.

Stephanie Ciccarelli:

Mm-hmm. ,

David Ciccarelli:

that's because the, that's the, the content is so dry.

Stephanie Ciccarelli:

Yeah,

David Ciccarelli:

But it's, it's 1200 lines, uh, along you have to read it verbatim and basically, you know, every word has, you know, a text word has a matching, and every line, uh, line of text has a matching audio file. And that's how the system, uh, is, is, is being trained. Now, the quality is going to be better because it's multiple hours. Um, but those are generally the two, the two approaches, uh, to getting a voice, uh, cloned.

Stephanie Ciccarelli:

Yeah. Well, that does explain a lot about the voices that you're hearing because Wow, like if they could have found something a little more uplifting, positive, happy <

David Ciccarelli:

mm-hmm. ,

Stephanie Ciccarelli:

um like, it, it just seems like, oh, what a text to read. Um, you know, when Well, I

David Ciccarelli:

Well, I think if, if we were to ever do this, and, you know, for those who might be interested, I mean, we're, we're certainly considering, um, entering into this space. Um, we would definitely recognize that there are, you know, not only different gender, uh, male and female voices, but also languages, you know, accents perhaps could, could play into it. But the key here is like, style and like the role, like who are you trying to perform and what's the style of the performance? And so rather than, you know, one long script of three to four to five hours, you know, might it be a better approach to say, ‘Hey, there's five different models for you that represent the five most common vocal archetypes that you perform most five most common styles. And reading a one hour sample of each of those, it probably gets you a more realistic sounding that could sound everything from energetic to, you know, an instruction, uh, an instructor, kind of a teacher type voice.’ So there's, um, that's kind of our thought of, you know, how, if, if anything, how we be able to do this differently.

Stephanie Ciccarelli:

Yeah. I'm all for a different script. I mean, that's gotta be such, oh, that's just, that's gruelling. Um, so anyway, I as we talk about this, uh, no doubt there's more to this story of where this is going, um, certainly more that, that you have to share. But, but David, is there anything else that you wanted to mention to our community now that you've got their ear about AI voice?

David Ciccarelli:

Well, I think that, um, you know, we are, we are active in the space behind the scenes. We are, uh, at this point, you know, have achieved a, a number of technical milestones with the ability to clone a voice, I would say, and put out, you know, a vision of what if a talent could have their voice cloned, you know, at, at really no cost. Um, and be able to put that available, you know, for hire, if you will, on voices.com? So make it so that hey, a talent, uh, you know, a a client could visit your profile, work with you, invite you to a job just like they always have, or, uh, if you've had your voice cloned, uh, make that available for hire as well too. For those times where the job's not interesting, it's really either in one extreme way too long or way too short, or it's an incredibly tight time, uh, turnaround time. There's lots of reasons why that might be an opportunity for, uh, for talent. So that's directionally what we're thinking about. We're really just getting started with this, uh, with this now. And, uh, I encourage people to, you know, stay subscribed to the blog, um, the podcast, uh, keep up to date on our email newsletters, uh, where they can learn more.

Stephanie Ciccarelli:

Wow, that is really interesting. Um, you know, just a thought that someone might be able to sell the clone of their voice on the site you had mentioned, you know, if it's not interesting as in, you know, it, it's just, it's just kind of stuff that needs to be recorded because it's, it's, you know, something, something that's not exciting, but, but still good work, work, you wouldn't have turned down. Um, now do you also foresee talent being able to, to, um, authorize the use of their AI voice, say if they're on vacation, like, I mean, this is no different from being sick or something and not able to record. So, would that work for them if they just hit a button and they say, yep, all done, and they could be on the beach wherever and, and their voice is working for them?

David Ciccarelli:

Mm-hmm. ? Yeah, I mean, I think the, what I'm understanding is from consulting with a number of talent already, um, just again, in, in conversation, is that this, this notion of, um, feeling like I have, uh, you know, that, that, that you have control over the words that are being generated by your clone voice. And so, you know, we kind of have this, you know, three step approach that we're thinking about. First off, um, that, you know, all if, if a client were to paste in a script into this kind of, you know, AI studio, if you will, that's, uh, that's, uh, on your profile, um, that it, that, that script's gonna go through. Of course, our, our content guidelines and our content guidelines are accessible, and the, the bottom, uh, we call the footer, the bottom of every page on our website, you can view them there, but in effect, they screen against, uh, a client script that may be unlawful, um, you know, uh, uh, unsuitable content. Um, and, and it kind of, you know, a number of, uh, trigger words, if you will, that we're like, that doesn't seem like appropriate. Um, and so that's what our content guidelines are there to screen for. That would certainly be, um, uh, deployed against any script that would be, you know, words attempted to be, you know, put into your, uh, into your mouth, in which case it just wouldn't even generate a preview. Client wouldn't be able to hear anything. The next up though, would be maybe there's some personal preferences or personal controls where, you know, you'd be able to screen against, you know, I'm, I'm supportive of this group and not that group, this political affiliation, not that one. Um, I like dogs, not cats. I know I'm being silly, but you kind of get the idea of like a number, a list of words that I don't want these words generated by my clone.

Stephanie Ciccarelli:

Mm-hmm.

David Ciccarelli:

So you'd have to list those out. Some talent of expressed concern or, uh, I frankly like that idea because maybe they have exclusivity arrangements where I can only work with, I can't work with any other airline, so therefore I have to put in a number of, you know, maybe I, if I put in a number of words that relate to airlines, you know, it, it, the AI voice wouldn't, uh, that that clone voice wouldn't be, um, generated that preview. And lastly, um, if those would be called it preventative, um, measures, then, then the detective measures like, okay, after the client generates the preview goes through and pays and downloads that, that you'd be able to gain visibility. You'd actually be able to, as a talent, uh, have access to see the words that were generated by your cloned voice. So preventing the bad stuff to be generated, and then giving the transparency and the visibility to actually see what was created afterwards. I think that, you know, two-pronged approach, um, should alleviate a lot of those concerns over what the words, uh, that are, that are being produced actually are.

Stephanie Ciccarelli:

Right. So that's really comforting, I think. Um, are there licensing agreements with these kind of voices too, that are put in place, say, on their profile if we go that way?

David Ciccarelli:

Yeah, yeah. The licensing, um, you know, for, I mean, the, the content that is most of interest here is really non-broadcast. Um, I don't picture both the quality, um, is, you know, at, at risk of being overly technical, like the download quality and what's being generated is, um, 22 kilohertz, not the 44.1. So right away, I mean, it's not, you know, you know, quote unquote broadcast quality audio. Um, so the use would just be probably very, you know, uh, industrial type, non-broadcast type uses, um, the client, just like any other licensing, it's really the client's responsibility to disclose how they're going to use the voice. So that would be something that we would ask. Um, and then, you know, make sure that there's some, uh, again, there is the idea around visibility, being able to see that, hey, yep, they said it's non-broadcast. Here's what it is. I'm listening, I'm reading the words and listening to the playback. It's clearly a non-broadcast, you know, um, application. Uh, and therefore, you know, it's the, the, the appropriate and commensurate licenses applied. That's, um, you know, again, the current thinking on this.

Stephanie Ciccarelli:

All right, well, I know that there's a lot left on the table and, and more that time will tell. So, uh, hopefully we can have you back on the show to talk about this, uh, when it's a little more developed and more, uh, voices involved. So, uh, thanks for coming on the show, David.

David Ciccarelli:

Yeah, I'd love to. Thanks Stephanie.

Stephanie Ciccarelli:

Well, ladies and gentlemen, that's the way we saw the world through the lens of voiceover this week. Thank you so much for listening to Vox Talk. Oh my goodness, the topic today. Such a hot topic, my goodness, AI voice, David Ciccarelli, CEO and co-founder of Voices here, giving you the scoop on what is going on in the industry. If you enjoyed this episode and you think someone else might enjoy it too, be sure to share it with them. Bring it toward your network. I know this is a big topic in voice over, so if you wanna hear what David has to say, get, um, sent out to even more people, then you know what your job is. Your job is to share because that is caring, right? So anyway, um, for Voices, I'm Stephanie Ciccarelli. You've been listening to Vox Talk. The show is produced by Geoff Bremner. Thanks again for tuning in, and we'll see you next week.