David Ciccarelli on the Future of Voices.com using AI voices
A series of recent events sent the VO industry into a spin last week, culminating in voices.com revealing that it would be aggressively expanding into the area of AI voice, TTS or speech synthesis.
Many reports circulated that voices.com, due to a change in terms of service, were now able to train any voice model using demos, auditions and jobs uploaded to the platform.
I put the concerns of the VO community to the CEO of voices.com, David Ciccarelli - and these are his direct responses. This content is provided for information only, and for the purposes of fact-checking so that interested parties can make up their own minds.
Here is a transcript of the interview:
Toby Ricketts
This is VO life hosted by me Toby Ricketts. It's a podcast, which goes over all the things happening in the voice over world. We meet the people behind the voices, and the people behind the companies that employ the voices. And today we have a very special guest, David Ciccarelli from voices.com. Recently, we've had some controversial news that's come out about voices.com, which is in the area of AI. So welcome, David. Thanks for coming on the program and answering some questions today. Of
David Ciccarelli
course. Great to be here, Toby, you and I go way back. So happy to to chime in.
Toby Ricketts
Absolutely cool. I thought Firstly, I want to do a bit of a recap bit of a timeline about the history of AI voices. And I don't like calling them AI voices. And I've heard people say it's not AI voice it's nothing AI about it. All it is is text to speech or speech synthesis. But we will probably mention AI voices from time to time. So anyway, back in 1960s, IBM developed speech synthesis, you know, you got the Stephen Hawking voice, you can hear what it's saying, but it doesn't really sound like a human. Then in the 2010s we've got machine learning Alexa, Siri, Google come on the scene in 2014 11 and 16, respectively, they offer like dynamic speech so that you know the computer can generate some text and read what's on screen or read you know, commands etc. Then in 2020s, the voice models improve Google's taka Tron two comes out in 2017. And lots of sites launch using this as kind of a face plate and basically just using the API to generate speech and then selling it to the public. And this is the first idea that as voices as professional voices, there's some you know, there's this stuff doing a very bad job of what we're doing. But in the last two years, there's been huge investment into like, you know, in data purchase and mining for speech data on the internet, machine learning generative AI comes along, of course, Chet be GPT is the most recent edition of this in terms of like generating text and just getting hyper realistic passing the Turing test and really smashing through all those barriers that we thought would 10 years away. Then there's stuff in the in the visual field like Dali, mid journey, stable diffusion comes along, starts, you know, copying famous works of art, extending famous works of art as the most amazing thing where they can basically paint what the the original artists didn't. And, and now it's kind of crept this photo realism is kind of starting to creep into the world of professional voice. We've got things like 11 Labs back, we've even got people being replaced, like with AI, radio, and DJs. You know, effectively fake these, as you can't tell, are not human. Then suddenly, last week, all of this stuff happened. We had a new song released from one of your fellow countrymen, Drake, one of the famous most famous rappers in the world, except it wasn't Drake and the weekend, it was a song that was completely contrived using AI. X next. And a very interesting sort of sudden court battle ensued, which we'll get into a little bit in the interview, but it proved that there were holes in the copyright law when it came to copying people's voices with AI. There was a Wall Street Journal article that came out at the beginning of last week, which outlines how lots of voice of artists and a lot of the people I know and work with, how they're being stolen from, you know, old samples that have been hoovered up from around the internet, or previous jobs that they've done for clients, where the terms of service were changed, or something happened around that. A good friend of mine and host of The Pro Audio suite, Andrew Peters announced on his podcast the other week that he's been replaced by AI for one of his biggest gigs, which has been the voice of a network. I got interviewed late last week about about how my voice had been stolen. And I was suddenly like, has it and I looked into Yes, it had I've been I've been copied on the internet. And then right into the middle of this whole fray like all this is going on. And then the voices.com announces the purchase of voices.ai. And and you updated your terms of service. And this is where everyone the internet went crazy for a second, lots of fires were started, pitchforks were grabbed and charged with and there was lots of lots of robust debate, shall we say around the internet about what it meant? What does this mean for me as voiceover, there's lots of existential crises going on with with with people's voice and their careers because like, you know, many people rely on this for their, for their bread that brings them to their house. So it mostly centers and this is my first question and mostly centered around the terms of service of voices.com. And we've covered this before in other discussions, because I remember when it changed, and it was like, you know, we own the voice for whatever we want all around the world in every jurisdiction. And that was correct me if I'm wrong, but that was basically so you can transfer the ownership from the voice to you to the client, so you can therefore Chase pregnant. That's it in a nutshell, right?
David Ciccarelli
Yep. So if the client doesn't pay for any reason that we have some recourse to go after the client to say listen, you don't own it. We own it as a temporary kind of holding pattern. On behalf of the talent until you pay Yes. And then we dispersed the payment to the talent. So that was that was the background on that project.
Toby Ricketts
That's advice. I assume you've had lawyers have got involved, and also only these rights otherwise, we can't do our jobs. So but what people are saying is that that effectively does like it puts in writing that you can use it for all these purposes. And while VoiceOver is not in, like explicitly mentioned, it's not explicitly excluded, I guess. So my biggest question is, like, you've made statements basically saying, We will not use your voice for AI, we won't use demos, we won't use auditions and we won't use and the end clients are not allowed to use Final jobs. And this specifically stated that it's an AI job. Why can't you? You've said that in statements like blog posts, and you've said it, you know, on interviews, etc. But couldn't you incorporate that into the terms of service? Because that would put a lot of mine people's mind at ease?
David Ciccarelli
I you know, and the short answer is, yeah, we're going to, and I think that's what we've realized that it was maybe an omission, through, you know, realizing that the technology has kind of developed that quickly, that this is needed. And it's why we co create with the talent community. And so we have some terminology with the lawyers right now, I was hoping that I could get it, you know, if you will, approved prior to prior to our podcast, but I'm happy to kind of read it out, because I think it's the section here, that would be most helpful that it does, quote, does not permit voices to make or trained derivative works. And then might be asking, well, what's the derivative work, such as synthetic voices, or more commonly known as AI voices of any user generated content without the express consent of said user, ie the talent? And so that's, that's what we have, I think it was just, it's really just the call out because derivative works. You know, it could be a cut down or an edit or something. That's how I think initially, we're thinking of derivative works not to create a whole synthetic voice. So that's what's that's what's kind of in queue right now. Yeah, there's, I mean, we've never created an AI voice. From, you know, from a from a demo, from an audition and the finished product, it's owned by the client, if the clients paid for it, that's their ownership at that point. And nor do we have any plans to so this wasn't, this wasn't a difficult decision to make. It was, as I say, perhaps, an oversight and omission, mostly because we didn't recognize the speed of the in the development of the technology. And if there's concern that we're doing this, I'd rather dispel the concern and put it put it in writing. And I appreciate that I can, I can make pledges and statements, but it's, it's really the legally binding agreement, which the community wants and happy to oblige.
Toby Ricketts
Okay, cool. Does that also cover the fact that if I, you know, say, I do a, you know, a three page elearning, for for Microsoft, that then they own those files, once they've, once they've paid for it, and they have this files in their position, it's very easy for them to train something like a living labs on that on just, you know, three minutes or 10 minutes of audio, and to sound like my voice does anything in the terms of service prevent clients currently from using it for a different purpose than was hired for?
David Ciccarelli
It does similarly, how, and I'd have to kind of dig up the exact kind of section number, if you will. But certainly the intent is just like if you're going to have it for an audiobook, and that's presumably hours long, it's for an audiobook. It's not meant to train another system. And maybe in a more kind of practical use, you said, it's for a radio commercial, you don't get to transpose that and put it into an online audio ad and upload it to Spotify, and get, you know, a million listens that way. So this really does come down to the honor system of the client saying this is how I'm going to use it. Our responsibility is to ask the right questions, and capture what they're saying, in terms of usage, and ensuring that we communicate that through this system to the talent and let the talent decide. Yep, that's acceptable in terms of the use and the price. That's what I'm agreeing to. And yet, I think we can go a step further. And so next week, you know, there's there's a lot of these licenses that have been established. I've granted permission for my voice to be used for this medium for the next time period, 13 weeks or a year. And yet there isn't, you know, a log, if you will, a system and we move asked talent, how do you keep track of all that usage? And there's a phrase in the industry that the real money is made on the back end, it's not actually the initial particularly in broadcast work. It's not necessarily the initial creation. It's the ongoing usage of that for real successful campaigns that might go on for multiple years. And so we wanted to build a in what we're launching next week is a license management system, where the client To be able to see all the licenses that they've purchased. When it comes up for renewal, we'd ask them, I think it's like 30 days in advance. Are you still using this? Do you plan on using it? If so, would you like to renew your license? And then there's a request for quote, and so that they can ask the talent, hey, what would it cost to renew this license for, you know, and they can extend the time period. Or they can extend again, modified, perhaps the jurisdiction of the media, if you will. So it is a request for quote, but it's, it doesn't require an audition, and it doesn't require the delivery of another file. But that at least creates the system for that kind of ongoing usage, to be, you know, managed and monitor over
Toby Ricketts
time license extension. Exactly. It'd be lovely to see some kind of collaboration with eyespot. TV or one of the media monitoring platforms as well, so that it can actually, you know, because I know many people who found work that are still playing that should absolutely, so that's, that's exciting progress. I'm just going back to the Terms of Service, again, in terms of what because many people were, they were like, it happened on the same day announcing the AI voices and then changing the terms of service. And of course, someone who's, who's a bit worried about the career anyway, will leap to the conclusion that you change Terms of Service, so that you can enable you to train voice models for whatever, whatever, you've gone on the record saying that that obviously didn't happen. It was a kind of a minor change. I think they happen on the 24th do want to just outline people who, who didn't see the explanation for that.
David Ciccarelli
Sure. No, thanks. Thanks, Toby. The change on to the Terms of Service last week actually related to we've removed something from the terms of service, which related to language around the categories that we, you know, moved into, you know, we've always we've been doing for voiceover for more than 15 years, a couple years ago, we thought, Oh, well, you know, audio production is a natural extension, maybe music composition, or kind of custom music. And then translation, often scripts need to translate from English to French and Spanish and so forth. And so why don't we enable these additional creative services on voices.com, then we had lots of, you know, service providers, other talent in those other creative categories. But really getting the clients to think to go to voices.com. First for translation was a real struggle, even though they had, you know, maybe done some work with us in the past. So that was
Toby Ricketts
an experiment that you launched them didn't really work,
David Ciccarelli
exactly, we looked at, and I think we're a niche player and, and really excel at that. And I think that's something that we're recognizing, and so we should, you know, in all likelihood, embrace the strength of being a leader in voice. And you're right, it candidly didn't work as, as anticipated. So we thought, why maintain these because every, every new feature and iteration, we talked about whether it's licensing, or, you know, searching the search engine, we always have to ask, Well, how do we do that for translation? How do we do it for music, and it just slowed us down? Okay, so we decided to wind it down, remove the language, you're right. Coincidentally, had these are two completely unrelated items. Now, when we make a change to the Terms of Service, there's, there's three levels, okay. One is this fundamentally changes the nature of the agreement, the spirit of the agreement, if you will, between voices, and the clients or the talent, in which case, and that could be around pricing change, you know, cancellation clause refund policy, these kind of big pieces. And in that case, you've probably experienced on some sites, you'll land and it requires you to click or accept the agreement before you even continue. So one, this was far from that this was kind of saying this didn't work, we've removed this language, one level below that would say, Okay, we're going to inform you that something changed. And we'll describe kind of what it is. But in this situation, you know, and that's kind of like an email goes out, and people are informed. But you don't need to click to accept it just kind of breeze breeze through the lowest level would be it could be everything from a typo, to, you know, removing something innocuous that we're just not doing anymore, or a change of address or something that just needs to be a, you know, an update. Still, we go through that goes through kind of a, you know, a past with the lawyers internally and externally. And then we publish the new terms of service and archive, I think, which is above the standard of care archive, the previous version. So that's, that was really the level of change. In this in this Terms of Service, the sections that have been mentioned, they've actually been in there for over two years. So it's not like we added this new section. At that time. It's been there for two years.
Toby Ricketts
We did an interview on it at the time, actually, which I'll link in this video to prove that that's the case. Yeah, it's been a while the other kind of couple of legal questions while we're on the legal side. Which does hear where this exists is like, When will this new wording be added to the Terms of Service? People want to know that? Yeah, I
David Ciccarelli
listen if I could have it happened today? I certainly would. I mean, it went to them on Thursday, which was the April 28 or so. It's usually a week turnaround. It sounds you know, innocuous, just to add a couple words of like, trying to be more specific. But, you know, you know, the fact that we're even having this conversation shows that every word matters. And so just want to make sure that we've, we've gone through the process there. But because this is the commitment in saying, we're not going to do something this would, you know, be at that level of, of sending out an email, informing the community and describing in that email, Hey, here's, here's what changed. And here's, here's what's new and different as of this date.
Toby Ricketts
Okay, cool. And on the low question as well, someone brought up on one of the forums that, that you're based in Canada, and it's Canadian law, which governs your your terms of service, and which, which doesn't have any statutory performance rights provisions like in the US, and produces utilize work for hire doctrine there. And apparently isn't much so that we can do about it. This is this is the problem that's being seen around the world in various jurisdictions where, like, for example, that the Drake video where they found that they there wasn't actually any law to be able to prosecute this person, because then they've done anything wrong, according to the law. So is there is it something you've thought about in terms of like, is it always going to be governed by Canadian law where there's not as strong a copyright provisions in California law, for example? Like, how does that implement your your terms of service and your business model? Do you think we can sort of get what power clients have in this new AI world?
David Ciccarelli
I don't think it weakens it. I mean, we, as a Canadian entity, we do, you know, abide by Canadian law. The the new site, which again, we're really just standing up, is not another legal entity. voices.ai is really just a website at this point. It's not it's not a distinct legal entity. So it certainly would fall within Canadian law. And, you know, in the in the United States, the actual jurisdiction where most of the corporate work where a company maintains its corporate headquarters or entity is actually in Delaware, because there's so much precedent for case law in Delaware. So each state might operate differently. But if it's, you know, it Delaware has like the history and there's something unique about that space, it's not a very populous state at all,
Toby Ricketts
I had a podcast about that the other day about how it is like, it's all of the corporate laws are written by corporate lawyers. And literally, it's like, entirely closed off process. So it's basically like, you can do whatever you want in Delaware.
David Ciccarelli
That's, and that's where you if you were to incorporate a new business, it almost defaults to being in Delaware. Yeah, that's where you're Incorporated. You can operate really anywhere with with your office. But we're Canadian company. We've not viewed this as a hindrance in the past, and certainly hope not in the future, either.
Toby Ricketts
Okay, cool. That's so good. I guess just to round off that decision about because still people will have an issue with, you know, the terms of service that have been enforced for two years saying, you know, we can we can use, we can use this for whatever we want. There's Do you see there is any other way around that in terms of like being able to be the intermediary between voices and clients, you will always need that permission, if just expressly for the purposes of completing a commercial transaction? Is that your your official view? And there's there's sort of no other way? Because that does put a lot of people off, even though it's probably on other sites to be honest.
David Ciccarelli
Yeah, I mean, we've, I guess, in the absence of that, we it really doesn't give us any teeth to go after the clients, like how are we engaging the clients? I mean, I think the key difference is that we run a transactional platform, the money and the files are delivered through voices.com. Most from the best of my knowledge, most other voice kind of casting sites, really act is purely the matchmaker and say, Okay, go off to PayPal to figure out payment and send files by Dropbox or upload them to Google Drive. They're not handling the end to end. And so because we are, which is actually our patent that we hold in the US and Canada, which you know, we pervert, I guess properly referred to as sharpei. But there's a patent that's held on that. It really does require us to have that kind of end to end kind of coverage. And why which is why the language is a lot stronger there because we're facility transaction from the get go right through file from payment and file delivery as well.
Toby Ricketts
Interesting. It'll be interesting. If was 123 has changed, because they had been now, you know, trying to get as many talent through their platform on their, you know, them organizing the payment sites or their terms of service have changed. But I get back to our main topic. So voices.ai, you bought voices.ai? Was Was this? Was it just a name? Because, you know, I went to the site expecting demos like on speech alone there. I mean, there are 1001 ai voice sites now, because everyone wants to be the new, you know, center of artificial voice. But I was kind of disappointed in terms of it had lots of lots of statements like all the others and pictures, but you couldn't press play anywhere and hear the results, you could sign up for updates. Was it basically just a park domain that you bought? Or did you buy any technology along with that? Or any kind of data sets or anything like that?
David Ciccarelli
Yeah, no, I'm glad you brought that up. Because it was it was, it was really just a domain. Originally, it was just purely to be a defensive move. I mean, I wanted to acquire the domain, so that it didn't get into the wrong hands, ie a competitor, or maybe a new startup causes all kinds of confusion of, you know, hey, is this the site kind of looks and feels like you, there's actually even some quite some sites nowadays that are that are just voices, and then some something else, and people think that it's us. And so that I kind of saw that playing out. And I was like, well, maybe we should, you know, acquire this name, as the kind of ideal, I did what most people do, you just type it into the browser at the time, it was just simply a parked domain, which, which for those who are unfamiliar with this term of park domain, it means a website that is just hosted by the registrar, it's kind of a placeholder page, and it says, you know, website coming soon, something like that, and some contact information for the registrar, there was no, there was no site on it, there was no company or technology. In so after speaking with clients, though, you know, being someone I try to connect the dots on this type of thing, we've actually identified a new type of customer, which is that of the software developer, you know, on voices.com, mostly, it's creative producers, it's video producers, it's brand managers, marketing agencies, and, of course, small business owners as well, too. But this new persona, if you will, is that of the software developer, and these are people who were wanting to incorporate, you know, a synthetic voice into their product. And maybe it's for, you know, content that changes all the time. And therefore, you know, it's kind of a two way conversational assistant, or a chatbot. And so realizing and kind of identifying this new, you know, this new use case, if you will, and this new persona. At the same time, we were scheduled to attend an upcoming conference, on this very topic. I just felt, hey, why don't we soft launched this website, with this vision of actually creating a voice development platform. So a development platform is, you know, it's not really for the creative producer. It's somewhere where you get code snippets and documentation, and yes, through through code, you can access these type of AI generated voices. So if you go to, you know, Amazon Pali would be something for people to Google. It's the same type of voice development platform by Microsoft as yours manifest the same sort of thing. Exactly. Yeah. And so IP is basically it's really a tool set developers to create applications and seeing that, you know, chat right now, and is very popular, but it's really just tax base, that next evolution is going to be conversational, two way. And it's going to be voice based. And so I, you know, really just spotting an opportunity. So that's what this this site is there's you're right, there's no samples that you can play. It's really a join a waitlist, but we've been actually reaching out to those that have signed up and just said, Hey, what brought you to the site? Thanks for joining the waitlist, what is it that you're looking to build, because maybe we can help you sooner than later or when we co create with you, let's make sure that we're building the right thing. So there's 100 software developers on the list, and in certainly growing by a few every day, and you know, the future we're going to build with them as well, too.
Toby Ricketts
So you don't want to miss out on that. Because it's, it's clear that this is the direction that the world currently is heading in in terms of like heading voices. And you know, no one wants to miss that that piece of the pie, because we can see that it is going to be a big part. And I mean, we've been seeing it as voice artists for years, but didn't expect it to get this good this quickly. I mean, you'll be dragged into all of the debate that's raging at the moment, which is like it's large in the voice of Stockholm. It's mainly about copyright law. The Drake thing that I talked about earlier with the rapper, who you know released the song that he didn't actually sing at all, like they just sampled his voice, wrote a song and then wrapped it and then had their voice put over the top of it. And it highlighted all these these really interesting issues around copyright and the fact that like, No Copyright Act currently covers synthetic performances, just distributing copies of recordings. And so since AI voice recording is not a performance, given by the person who holds the copyright, it's a new, it's a new work independent. And so like lots of people are saying, well, this is, you know, the service. This basically allows people once they've got your voice likeness, as long as you're not performing it, and it's an AI performing it, then it's, it's, you know, the copyright belongs to the person who created the file so that in what ways I mean, this has been outside of the I don't know maybe what you've prepared for, but like, Have you identified ways to be a part of that debate? It's important now that we try and find a way that's equitable for voices to go forward and try and make some kind of living off off the synthetic voices they create. And you're going to be a big part of that, that discussion, given that you've got probably the one of the world's largest voice data sets, although you said you won't use it for for AI training? What are your thoughts around around? What's going to happen with copyrights? And how voices can keep control of their their voice?
David Ciccarelli
Yeah, well, to the best of my understanding, you really can't use someone's image or likeness, in for commercial purposes. Now that might be the difference is that that musical piece was done as almost like fan, a fan piece, it wasn't meant to generate revenue. Just like you can't take a picture of a celebrity, and then use it in an advertisement, right, without their permission, because you're using their image or likeness. Now, this is why we're gonna say quite sensitive about this, when a client would ask for a sound alike, like, are you doing an impression of someone? Or are you trying to impersonate them, because if you're impersonating you are really trying to be that image and likeness of that talent. Classic example is James Earl Jones. People would say like, I want you to sound like James Earl Jones. So are you putting it off as being them? That's who you hired? Or is it an impression, where it's more satirical, if you will, in its in its performance? It's so I think there is a fine line. However, you know, the law clearly needs to be updated to handle this, but probably that distinction is if it's for commercial commercial purposes, on voices in particular. You know, how can we maybe prevent I think, is it might be where, what you're alluding to, like, prevent the misuse of, perhaps someone's cloned voice. You know, a couple things right now, we don't allow clients to, you know, download the auditions, or sorry, the demos, you know, it's really hard to extract, you know, you can't just go in and like, go to Google Images and get every image that's available. Or some of these sites where you can like download on mass. We, if we see kind of like, bought like activity, we're going to put, you know, we have firewalls, we're putting up blocks, it's quickly. So these are kind of preventative.
Toby Ricketts
Just just just just to expand that theory, there was a debate a long time ago, and it still kind of goes on a bit about watermarking. Because you know, and I was shied away from watermarking, because, you know, you'd only read, you know, most of the scripts or whatever, not include the brand name or something. And that would be that would prevent them from using it effectively. But now that we're talking about the actual likeness of a voice, is I mean, if everywhere, if every demo had a slight watermark in it, it would be quite useful because then someone couldn't just go and like write a script that would go and harvest every sample of a page. For example,
David Ciccarelli
I've heard that on. Audio Jungle has this little whisper audio chuckles from whisper in the background, a lot of sites do that. If you recall, I actually think it was voice 123 A couple years ago did that and clients were just, you know, talent, were like, you're, you're deteriorating or degrading my voice, I want this demo. And clients in the audition were like, this is kind of useless, you know, like they might want to play with at a team meeting or a collaborative session when they're making those decisions. So you know, I've witnessed this a few times. We've contemplated that for, like for auditions, it's just becomes incredibly distracting. I think most clients have good intention of they're going to use this audition so they can make a decision and ultimately hire a talent about it. So I'm not sure and when it comes to an AI generated voice, this is something else really important, I think for the community to know and understand. The highest fidelity that this can be done right now is 22 kilohertz, which is kind of half the quote unquote broadcast quality. So I am having a hard time believing that big brand out there who's going to spend millions of dollars on airtime and media buy is going to use a degraded low fidelity. I mean, it would be like, you know, audio file it'd be like seeing a billboard with it's all pixelated or like a watermark on on it.
Toby Ricketts
So I think the content is more than that. It'll be used for the long tail audio like, like like audiobooks, you know, elearning industrials, basically industrials is where is where it's at, in terms of this new voice technology, because that doesn't really require the acting and the quality level, most of its watched on YouTube on phones. So really, like the quality is not as much of a barrier as I thought, I feel like commercial, and commercials will be safe for some time yet.
David Ciccarelli
I would agree with that
Toby Ricketts
it's the train, it's kind of the training grounds of voice over the, the stuff that you do, when you're you first start out that is, seems to be mostly in jeopardy. And for that quality doesn't matter as much. Because that used to be my argument was just like, you know, the speech synthesis models, they only sample at, you know, 2205 hertz. So, it's, it sounds to me as an audio engineer, like, it's not very good quality, but I think people watching on their phones and on YouTube videos just don't don't even perceive that. Right, quality is really amazed at how quality is dropped. For even for things like commercials, sometimes when people are recording stuff on their phones, or through the Blue Yeti mic or, you know, and and it's kind of deemed acceptable. Now, we're absolutely wouldn't have been five years ago. Yeah, so that is a bit of a problem, you know, in terms of people that the bar to, to acceptable audio is much lower than it was I think, even with with regard to the performance, because you do hear some YouTube videos that are clearly AI voice generated quite badly. And yet, it says hundreds and 1000s of them. So
David Ciccarelli
yeah, I mean, the genres, if you will, that I would almost describe as like, the applications or the uses, where, you know, voiceover might be impacted, or kind of is like might tend towards an AI voice. And I know, we're using that term kind of interchangeably with synthetic voice as well, too. But I would look at the spectrum of like informational versus emotional. And if it's just pure information, like turn by turn directions, the elevator, the parking garage, the airport announcements, I mean, it's very, basically public service announcements. Or, it's, as you describe this kind of corporate training material, where the person in the curriculum departments, you know, and human within human resources, and they're the curriculum designer, I mean, their mandate is to make the content accessible to the most people possible in that organization. So that there's a new compliance or safety video that's being produced, then they need to make it available in five different languages. And the whole thing does needs to be, you know, so that people of all walks of life, whether you have a and perhaps even disabilities can access that content in a lot of ways. It's purely an accessibility play that is being done here. And they don't unfortunately, have a lot of budget for it. And so the tendency can be that it's 100,000 words, and we're just going to have the generated text to speech voice, do this, rather than it gets in sometimes it's not even economically feasible for the talent who's going to who's going to do 100,000 words, for a few 100 bucks, no one is going to bother. So they're in a bit of a conundrum themselves to abide by accessibility, whether it's a law or a mandate within their organization, but then aren't given the appropriate budget. So how do they make those to fit? But basically, it's informational content. That can often be very high word count. Or on the other end of the spectrum, very low word count. Still informational. But it's, it's five or six words, it's almost, and I hope this comes out the right way. But it's such a short shelf life that the content has, you know, it's a one time social media video that, you know, once you've seen it in the feed, you don't want to see that again. I've seen and heard that video before. vapor? Well, exactly. And a lot of a lot of clients are like, producing 10 different variations of that video to see which one sticks, right, different calls to action, different script even. And how do we do this at scale? There's one advertiser that actually produced 2000 They hired a talent to do this 2000 variations of an ad. And it could it was different price points. 495 496-490-7595, five different calls to action Sign up today get started today. And then language depending on the device and the location that was used. It was all this, this whole matrix of all these different combinations,
Toby Ricketts
dynamic ads, I think they call those now is probably the way there is I think as well like they had so they kind of construct them on the fly, which is a very you know, it's an Interesting new form of media. And that kind of combines, like, kind of it's gotten outside the realms of advertising. It's more market research now isn't that amazing? So
David Ciccarelli
you just and marketing, the phrase of her recently is marketing is math, you know, people are going to create all of these assets, they're going to figure out through the numbers, which one works the best with the most about people, and then run with that. It's not just, it's not merely a creative decision, we're going to create variations, and then pick the winner, so to speak.
Toby Ricketts
Cool. Again, like with this new, you know, building voice models, so you've kind of gone out and said, You're going to be creating voice models, this is with individuals on on the voices.com platform. So you're not talking about kind of creating faceless models that are generated sort of out of thin air or out of out of large data sets, you're talking about specific voice models that people will own, still on your platform, and then you'll be the licensing agent for that, is that correct? Almost like a reseller,
David Ciccarelli
right, it's still your voice, but we are reselling that voice on your behalf. So I think this is how we're going to do this differently than maybe as you describe kind of the dozens of other startups that are that are out there. They're trying to aggregate data to create an anonymous, I refer to them as the mash up voices, it's like five different voices all at the same time, you can't really tell who it is, there's no attribution back to the talent. So if you really wanted to work with them for the big national campaign, you couldn't, because you don't know who it is. So how we can do this differently is that inviting to start 10 men and 10 women to have their voices cloned on voices.com. And it's done by professional voice actors. That way, the client can always hire, you know, the natural professional voice actor. But if they have some of these situations, what we're hearing of like, scratch tracks, it's a spec read, or just trying to get a sense of the video timing, these type of situations, or it's a test campaign, and before we go invest in the bigger national or deploy the big media spend, let's just see if we can get some quick feedback from the market, if you will, kind of that market research idea. So I don't So in those situations, you you hire the clone. Now the ideal state is you start with the clone and kind of it's it's almost like an upsell to the professional. But some maybe it maybe they're just realized up that ad campaign didn't work. No, I would rather have that client, start that journey@voices.com and be able to kind of reuse the term upgrade to working with with the Pro, rather than going to two or three different sites and trying to, you know, figure all that out. So I think there's a number of between the kind of like that, that spec test work. And then this dynamic nature of content, I think those are two real promising kind of new forms of media that previously are the types of jobs that weren't coming to voices.
Toby Ricketts
So just for clarity, because at the beginning that you said 10 Men 10 Women are going to have their voices sampled, it sounded like you were gonna put all those together to make a clone, but you're talking about 10 separate clones, male and female, and they would read and people will want to know this, that the people whose voices are sampled are free at any time to stop using that, and you won't own that they still have ownership of their Yeah,
David Ciccarelli
that's correct. I mean, there's, I mean, again, there's what's in it for us to say, No, you must stay on the website, we're gonna continue to use this, I just think it breaks, you know, trust in violation of everything that we're trying to achieve, which is if you want to participate, here's a new and different opportunity. If for some reason it doesn't work out, pricing or timing, or for whatever reason is, like, Great, we'll just, let's call it, you know, turn off that service and availability on your profile. So, you know, easy, you know, no harm, no foul, easy way to, to take the exit on that one.
Toby Ricketts
So just wrapping up, because I know we've taken a lot of your time today and thank you for for answering the questions. Like, like, Why do you think I'm, I'm always surprises that you guys sort of make a change. And then suddenly, everyone just like goes crazy, is what is it about voices.com that people just seem to get so mad about I know that sort of there's there's a bit of history where things didn't go right in the past, you've addressed it and I feel like you know, you've you've you've fairly addressed that. Is it just that people won't like, like, once the trust is broken, they'll never trust you again, or what's your sort of strategy for trying to make people feel better about the platform?
David Ciccarelli
You know, this this, it's really challenging because we often do things first. And when you do things First, there's unknowns because there's no precedent, right? There's not someone who's done this, you know, five times or you know, and in which case, those who are gonna say, the best I can do is communicate, be here on podcasts and interviews and videos, maintain a blog, let people know what we're up to. But there's actually a few venues that maybe others are unaware of, you know, I certainly don't operate in isolation. First and foremost, we have a great board of directors, who all of these type of decisions are run by that include a couple of independent board members, as well as our investors in so they're, you know, our intentions and incentives are aligned to make sure we're doing right by the business as well as the customers in which we serve. I also, you know, consult, if you will, with a CEO advisory group. So these are 20 voice actors, that from time to time, I'll preview ideas to say, hey, directionally, this is where we're gonna go. Or this is a decision we haven't made yet, but I wanted to bounce it off you. And those are one on one sessions that are held. And in addition to that, I've been hosting a number of what I call voices local, which are really just private events, you know, if I'm in a city or traveling to meet with clients, or again, at a board meeting, you know, gather people around for dinner, and we can go deep, these often go for three, four or five hours at a time. And, you know, there's, there's, there's nothing, you know, nothing off limits. It's a you know, and there's also no agenda for the evening. So anything goes often it's, it's definitely talking shot the whole night, and I try to answer questions, and maybe provide some clarity where possible. And that's really been, I feel, Toby, it's actually been quite helpful in the last couple of years, since putting in a real effort to be seen, and hopefully be heard. So, you know, yet another commitment is to continue up with those practices, because I do think they're making a difference.
Toby Ricketts
Cool. But David says early toward coming to a city near you. One final question, will there always be a place for human voices who own their own voices on your platforms, and you won't, you know, exploit them as it were, like,
David Ciccarelli
Oh, my goodness, I think this I think, the futures people because people buy and, and like to engage with other people, that we dragged some research around, like, well, what are the most popular voices, and it ended up being that people like to, you know, in terms of advertisements, like to buy from people who sound like them, which makes perfect sense, right? It's whether it's demographically or geographically in so that he's just going to need, you know, the human elements. There's things that are very difficult to replicate, if you will, like, and I know, we just say emotion is a bit of a catch all. But there's a lot that's, that can be unpacked there. Things like timing or comedic timing, or pause, what's the right amount of pause that it sounds natural, or even just a breath? You know, I think we tend to edit a lot of breath. So sometimes you need the breath or the sigh. And so I think there's absolutely a place for human voices on voices.com, it's likely going to be probably 99% of the activity for for quite a while because clients say yet well, I've looked at it, I've tried it, I can see some situations here and there. And maybe they get started with it. But I think there's continues to be a bright future for talent for years to come.
Toby Ricketts
Fantastic. Well, thank you very much for representing today and talking about the controversy. And I think we we got, you know, we've got through a good chunk of stuff there anything else that that you feel like people should know about your your current moves?
David Ciccarelli
No. Well, I think if anything, know that this, you know, we have good intent, we are here in in collaboration with the community. It might be cliche, but I've said this many times, our business truly is based upon shared success with the talent. When you thrive, that's when we do which is why we create the content, which is why we bring in the clients. So we were enthused, and we want to continue to work collaboratively with the community.
Toby Ricketts
And if people disagree with that, and think that those aren't your intentions, is there a way for them to get in touch? Like, what's the best way to actually
David Ciccarelli
I would love nothing more than people to send me an email, my personal emails, david@voices.com, it's pretty easy one to remember. And no matter how you feel, or what you're thinking, or you see something or hear something that doesn't sound quite right, I would welcome that, because that actually opens up the dialogue. What is what is heartbreaking is the assumptions that kind of lead to you know, and that are that are not really based upon fact, or reality, or assume kind of the worst, that becomes very challenging to navigate. And there's nothing more that I'd love than just to open a dialogue and let me hear the other side. And maybe it's an opportunity for us to improve or minimally course correct. So I would encourage that if you don't, if you don't feel that way, simply send me an email. Let's have a conversation.
Toby Ricketts
And I mean, it's true. I've sent you a number of jobs. where I've been like, Oh, this looks really dodgy. This is not a good look and you've looked into it and you've either stopped it or you've explained the situation and it has it is a good dialogue and like there was a there was a job about this AI stuff that we didn't get into on this time. But you know, it's yeah, if you have any concerns, then then email David and you must get a lot of email though. That email address
David Ciccarelli
I do. It's it's a little it's a little scary, but I chip away at it. I spend a lot of time just chipping away at but so please extend some patience and some grace but I'll get you an answer.
Toby Ricketts
Dave. It's it's really CEO of voices.com Thank you for fronting today and appearing here on vo life we will be sure to be in touch. If there's any other questions that the community wants answered.
David Ciccarelli
You got it thanks, Toby.
