ZOO Digital Group plc (ZOO) Earnings Call Transcript & Summary

AI is transforming the global entertainment landscape. And ZOO Digital, given its roots as the technology and software business is cementing its position as an industry innovator. I'm joined today by Stuart Green, CEO of ZOO Digital, Chris Oakley, Chief Technology Officer; and my colleague, Gareth Evans, Managing Director of Progressive Equity Research.

Stuart, would you mind to give us a brief overview of ZOO Digital and in terms of what you do and how AI has been relevant so far?

Sure. Well, what we do is we work in the entertainment industry. We work for large clients who are producers and distributors through streaming services of scripted TV shows, feature films and so on. And what we do is take their content and prepare it, so it can be distributed globally on streaming platforms. And a big part of that is to adapt it into different languages. And you do that, of course, in entertainment industry using subtitles, where you put text over the top of the picture and by dubbing, where you actually replace the voices of the actors who are speaking. These are all areas, obviously, where AI has come forward quite recently as a possible solution for adapting these materials to produce different language versions and therefore, potentially to massively reduce the cost of what it would -- what would normally be required to produce those different language versions. So we've obviously -- we're an innovator in our industry. We come from a tech background. We've been creating software solutions for many years, and we're kind of at the forefront of this technology, looking very close to it to really understand what it can do, and importantly, what it can't do or where is limitations are. And our aim is obviously to capitalize on it where it makes sense, and it stacks up. But also to be mindful of where its limitations are so that we can kind of work around those and ensure that we're using it to its full potential and which we ultimately believe is a combination of humans doing what humans do best in creative pursuits in conjunction with the technology.

Just on that, Chris, being the Chief Technology Officer, can you talk a bit more about what AI means for ZOO and your team in particular?

Yes. So we've been looking at AI in terms of how it can potentially benefit us and be part of what the process that we're involved in. So we've been using -- you look at utilizing it as an assistive technology for existing workflows that we already have. So where we can either reduce the amount of time that it takes to produce content, where we can increase quality or we can reduce cost. So we've been looking at how that can be incorporated into what we do.

And in particular, for localization, can you say a bit more about how it can be used and what maybe some of the limitations are?

So localization of TV and movie content is quite different from localization of text-based content or something a bit more static. So it's got a lot of idiomatic expressions, is very kind of emotion within the dialogue, and it can be very different to, as I say, to existing text-based translation. So this is quite an art form around the creation of localized content. And the people that we work with to create this content have to understand what's going on within the film, what are the messages that they're trying to convey at that time. And then translate into the localized language that they're producing at that time. So we've been looking at how AI can be in addition to that and to apply into that particular scenario. It can't do these things right now. It can't look at idiomatic expressions, it can't convey the emotion, the complexity of dialogue yet. So we're looking at how it can potentially be utilized to be assistive in that particular area.

I think we've seen a real shift in recent months with more and more companies coming out and saying that they view AI more as an assistive technology rather than a complete replacement for humans. And the phrase is often used, AI isn't there yet, particularly for media localization and premium content. What's the perspective on this and how far away is yet?

That's a great question. I think I'll put it -- begin that by just maybe a little bit more color to what Chris was saying there in terms of what localization is in our world. And as Chris explained, it's about understanding the intention. It's not just about the words. So if you're translating an instruction manually, it's just a factual piece of information you need to convey in a different language, and that's very straightforward. But entertainment content is, by definition, intended to entertain. And so it's really about understanding the intention behind in those words spoken. And to do that, you need to understand so much more than just the verbatim dialogue as it was delivered in the original language. You need to understand what are the motivations of the characters. What is their background? Where do they come from? In what areas they set? What is the relationship between the different people on the screen, to understand the subtleties like the idiomatic language and the culture references and all those things. So to be able to do this, with an AI system or indeed any automated process, there is so much more that you need to be able to understand and interpret and make sense of, in order to be able to then automate this process. And in that sense, just thinking about the current generation of AI technology and how that works. Essentially, AI is a glorified predictive text. I mean it really is very clever software and very capable at finding patterns in huge sources of data and therefore, be able to take the average of those and then reapply them in a different area. So even when we think about subtitles, it's so important that those subtitles are right that they resonate with that target audience. And you only have to think about some of the reasons perhaps why you wouldn't enjoy a TV program. The subject matter, the plot may interest you, but there's something about that program that just fall short. The fails in that suspension disbelief that we all want to get from entertainment. We want to be at the moment. Even though what we're seeing and hearing on screen, we know it's not real, and we know it may be completely fanciful, dragons or something, and yet we can be held in that moment of suspended disbelief because of the authenticity of the whole thing. It's about how the characters are portrayed, how do they deliver the lines, how well those actors act. And when we think about localization, it is mainly an extension of all of that. It's about portraying that all those things to an audience that doesn't speak the original language of the program in a way that is totally authentic and believable to them, even though the subject matter may be completely fanciful. So in answer to your question, why hasn't it happened yet? It's because, as Chris was saying, there is so much to this. There's so much involved, there is so much you need to take into account. And at the moment, AI-driven purely by the text of what has been said, is missing all this information that really comes from a lifetime of experiences on the part of writers and localizers and so on. So in answer to your question yet. Well, I think there is a fundamental limitation, a glass ceiling, in what current generation of AI technologies can accomplish simply because of what they're doing is finding patterns in large source of data and then replicating those in a new situation. And therefore, I don't think in the current generation, it will be possible for those systems to displace the kind of creative pursuits that are so much a part of what we do as a business and what are many thousands of collaborators who are experienced translators and voice actors and so on are able to deliver.

Just on that, Chris, AI isn't really new for ZOO. You've been using it for some time, pre the hype when that ChatGPT, everyone suddenly heard of AI. And probably it's at the forefront of your business given your software roots. But can you say how you're using AI and your technology-led approach to build on this further?

Yes. So we have a research team, made up of software engineers and data scientists. And it's been their responsibility for a long time to look at how we can develop emerging technologies into the existing technologies that we develop in-house. So we've been working with it for a number of years, 5, 6 years now, I think, in terms of actually incorporating into our workflows and into our tool set. So initially, it was very, very simple things that just very assistive technologies to identify languages or to identify certain elements within the dialogue or to provide some QC elements to the content. But over the years, we've developed more and more complex systems that gives us a lot more to what we're offering within our tool set. Recently, we released a tool that allows us to do AI transcription. So we can take an original video we can run it through our tool, and we can provide an automated transcription that will segment all the content into the correct sections, and then it will identify who is speaking at each point. So it really gives us a huge benefit there where it would be quite a laborious process originally, to now be able to automate that and reduce that time down to be able to do it really quickly. And that improves our efficiency, but also allows us to scale more as well. So that's where we're at right now in terms of technologies for the R&D team and where we're looking at incorporating in those within our tool sets.

So just back -- I guess, back to the -- why isn't it there yet? I think this was a seismic shift to use AI, especially for premium content, just hasn't really happened as fast as expected. Particularly in the context of a like-for-like replacement of the creative processes, is there anything else that we need to be aware of to help us understand sort of the complexities of localization and entertainment, where it can be used, where it can't be used, just building on what you said in your...

Yes, absolutely. Well, I guess, in my explanation, I was really thinking primarily about ZOO's business and what we do and the kind of companies that we provide services to and what's important to them. And those -- and our customers are big brands in entertainment who spend a lot of money producing fabulous scripted TV series and feature films and so on. And for them, what's -- the most important thing is the quality and authenticity of the work that we do. It has got to be as believable in the target language that we're working on, or languages, as it is in its original language. And that's the area that I elaborated a moment ago is where the current generation of AI technology isn't up to that and where really you need experienced people to do that work. But of course, there's plenty of other content out there where those kinds of approaches are actually perfectly fine. So certainly lower-value content or content is very cheap and quick to produce. So think of things like user-generated content on YouTube. If you compute something very quick and very cost effectively, then the accuracy of the subtitles for example, is not such an important consideration. And if you can produce those subtitles very quickly and cost effectively using some AI system, then, of course, it makes sense to do that. So I think in the broader industry, there are certainly many areas where these technologies actually, even now, are viable and data are already being used. But at our end of the market, where quality is the most important thing, not so much. And the second is probably worth saying, the second most important thing to our customers these days, is about time to market. As soon as they've made these programs, they want to distribute them globally as quickly as possible. And so anything that we can do and the kind of technologies that Chris was describing there, they all help in our efforts to reduce the turn time for each project. And just to make it that a little bit quicker, so that we can serve our customers with the results of our work sooner than they would otherwise have expected them.

Chris, just going back to your point on using AI for transcription work. If we look at subtitling, I think that's an area that a number of investors see as maybe a low-hanging fruit for AI. But I was interested that you don't even use it as a first part for your subtitling work. Can you explain how you use it within subtitling and how you anticipate it will evolve a bit further?

Yes. So when we get a project delivered to us, we take the original video and we transcribed it and traditionally, that was done by a human who would sit there watch the video and type out basically what's being said on screen, time the text in line with the video and segment that up into chunks to be able to be read on screen. That process now can be done utilizing AI. So we can get a very close transcription out of it, and that just requires a human to just come and provide a final pass to us to -- just to fix any problems that it may have found and just identify any overlapping areas or problematic areas that will cause problems further down the line. What we found is we've run quite a few significant tests with leading models and large language models and we produced manually translated -- or AI translated content. And what we found is that there's such a huge amount of edits that are needed to be made by humans at that point that it's actually quicker to start from scratch than it is to fix that up. Mainly because it is translating word-for-word, it's doing a document level rather than actually understanding what the dialogue is needing to say, how it should be split, what the context is and what they're trying to get across on screen. So because of that, a translator has to go in and make a huge amount of edits to that and it does take a lot longer to do that. That's not to say that machine translation doesn't have a place, as Stuart mentioned, in certain types of content where it's more static, less expressive, voice over work, documentaries and things like that. It gets -- it more -- it can produce a more accurate result at that point. But with this script-heavy content that we work on a lot of the time, we really find it's easier to start from scratch. But in future, we can see where there will be benefits of utilizing both there, and they'll get to a point where we can potentially -- we can start from a machine translation point and work from that and that will be quicker in that approach.

And I think it's pretty interesting the kind of things that we've looked at to provide assistance to translators. So not to do the job for them, but actually to provide them with hints and suggestions and identify, for example, idioms is a key area because obviously somebody who doesn't speak source language as their first language, identifying what is and isn't an idiom. It's not an obvious thing.

Yes. I'm trying to -- you can utilize AI to identify what the context of is going on, on the screen. So I understand there is an emotional scene, that it's -- the people are upset. That can be conveyed too with the translator in the form of additional data that they can go alongside. So it allows them to produce better translations. And then it doesn't have a long-winded pass where it can be rejected because it's not conveying the right thing. So you can have benefits throughout the process by utilizing AI, but not necessarily just translating.

Okay. it'd be quite useful to sort of understand some of the distinctions between the technologies being used. We've seen some announcements from big names like Warner Bros., Google using AI for live and near-live broadcast for captioning. It'd be Useful to sort of understand your take on this and the viability of the sort of solutions.

It's one of the main things for live and near-live subtitling is speed. They need to get that our on screen as soon as possible. So existing way of doing that is respeaking. So they will have somebody with a microphone that will listen along and speak that back through to the computer, and then that will provide a subtitle stream. And that in itself is quite problematic if you've ever watched a news program with subtitles on a live news program. You can tell it's often quite far behind. There's mistakes that are happening. So there is a potential to utilize ASR, speech recognition in that context because that can provide a similar level of quality. But it still has its limitations. It still have -- it will identify things correctly, there'll be background noise that affects the ASR. So as there are impacts of potentially utilizing that as well. So it's found a place then that kind of goes back to the type of content that it can work with. It definitely has its place in that certain type of content. And that, again, is something that we're looking at as well as tools that can provide live and near-live subtitling, utilizing what we've already created as part of our workflow.

So we've looked quite a bit at the subtitling side, but dubbing, that's a slightly more complex challenge as I understand it with AI. Can you explain a bit -- in a bit more detail what the challenges are and what you see happening within dubbing and the trends that we're starting to see? And if -- and I guess importantly, the future for dubbing because it is quite an important part of entertainment.

Sure. Well, where to start. I mean, dubbing is quite a big area, isn't it? And dubbing so far, as you say, mostly about subtitling, which is really about getting the script right, the actual words that are said. When it comes to dubbing, clearly, you've got to -- you wouldn't deliver those lines in a way that is convincing and authentic and true to the characters and so on. So I think when folks talk about AI dubbing and you do a search for that insight, you'll find a whole bunch of companies that are offering solutions for that. I think what people are thinking of is a way to somehow go from the original program, in its original language, to producing a fully dubbed version in a completely different language. So really fully automating that process. And again, that's something there is technology out there that does that. And I think if your source material is cheap to produce, user-generated content, then I'm sure it's perfectly fit for purpose. But again, in our world, at that high end of the industry, the quality and authenticity are everything. And of course, there is so much to delivering the lines in a way that is convincing, being able to somehow infer the intent behind the lines, which is what an AI system would need to do in order to be able to then figure out how to deliver those lines for a different culture, speaking a different language, that's a really tough ask. So certainly, though, there are aspects of dubbing where we see already some really interesting and promising opportunities, and the teams are working on some of these for a while now. And...

Yes. So we've been looking at, not synthetic voices but looking at voice cloning and how we can utilize AI voice is to actually expand what a single person can do or what a single actor do. And one of the particular areas we've been focusing in is around children's voices, which are a very complex process to capture children's voices. They're limited in terms of the amount of time that they can spend actually recording. They're not like a human actor. So this complexity is there. There's issues around children's voices changing as they get older and you may want to record over a series and you may not get the same voice at the end of the series that you had at the start. So there's a huge amount of complexities there and that's one of a big challenge area that we decided to have a look at is how we can potentially take an adult human voice and convert that into a child's voice. And so now you can quite quickly produce child-like content without the need of hiring children. So as Stuart mentioned, there's a lot of potential there to incorporate AI in there, but it is a challenging environment. There's a lot of legislation at the moment, and there's a lot of contractual obligations that studios have to go through as well. So it's one that we're pursuing, but it's still a bit of a challenge area for us.

So do you think that some areas of dubbing will become more synthetic for a while, perhaps not in the premium content area you operate in?

I'm sure they will. I'm sure in YouTube, you'll hear lots of -- well, maybe you don't listen to French, German, the programs but I think in those other countries where countries different from the ones in which content is produced, you'll start to hear sort of voice over time dubbing used because it's a great -- it will be a great and very cost-effective way for a YouTube artist, creator to get their content to a wider audience. I think it absolutely stacks up there and makes sense and has a lot of mileage. But it's just a very different use case from the world in which we operate.

So I think there's quite a balance to be struck between the cost and quality and localization, particularly for premium content where quality is critical. But how is the -- and what are the opportunities for ZOO around AI to help with your traditional processes becoming more efficient, driving quality? And as you mentioned earlier, reducing the turnaround time?

I mean, I think there's a whole host of things that we've been looking at and working on some of which we've talked about so far in the session. But in fact, we haven't -- I mean we've only talked about media localization, but actually, that's just 1/2 of what we do as a business because as I explained earlier, we take content and ensure that it can be distributed globally on streaming services and for different cultures. But a big part of that is actually a technical, there's a whole range of technical things that need to be done to make sure the content can work effectively. And there's a whole host of processes behind the scenes that I won't elaborate on but where traditionally you would have people doing those things. And yet, we've -- it's been -- we've driven throughout the 20-plus years, we've been in this industry to automate the things that we do so that creative people can really focus on what they do best and what humans do best, which is being creative. And other things that are very repetitive and time-consuming and error-prone wherever possible, we always strive to provide systems that can do that work. And in the past, we've done -- used all sort of automation approaches and more recently, we're turning to AI as a way for us to potentially do some of these things in a very efficient way.

So just looking at what the major studios are currently doing in terms of outsourcing the process? Maybe Gareth, you can touch on this. I think the is benefiting from further supply rationalization, but do you see a threat from AI that studios bring the process in-house and looking at perhaps other technology stocks? Is this something that's happening with the more process-led technologies that we're seeing?

Sure. I mean the decision around outsourcing and technology is always a vibrant one, it evolves and it moves over time. And technology never stand still always evolves and moves forward rapidly. And I think AI is just the most recent and latest and probably most talked about iteration of that. But if you think about any large business, almost any business of scale at the moment is using AI -- looking to use AI in some way within their processes, and a couple of recent examples from companies in slightly different sectors, but for whom it's been quite a significant shift. One example is Klarna, the payments business. they're using AI. And they're saying they said recently that it's replaced 700 people within that business. The work of 700 people is being done by their AI platforms. And what they've seen is, they've seen resolution times for problems reduced from 11 minutes to 2 minutes so that the customers are getting a better experience, and they're saving a huge amount of money. ASOS, the online retailer, they've said that they're seeing significantly lower levels of returns because they're using AI to help customers choose the right products in the first place. So they're using it structurally within their business to make a material difference. And I think if we look at the localization space, it's quite different because those businesses, what they're looking for, they're looking for patterns to be recognized. They're looking for standardization. They're looking for simplicity, taking difficult complex inputs and coming in with one potentially right answer in each case. But I think localization is very different. It's a much more creative process. It's much more nuanced. There isn't one right answer. There are lots of different ways of doing these things. And it's a very important -- I think that's a very important perspective within the localization industry, which is quite different to other sort of structural uses of AI or ways that large businesses are using those sorts of technologies. And I think if you're trying to compare and contrast the localization space with those other sort of much more structural, genuinely business driving choices, I think there are 3 things. I think is cost, its complexity and its creativity. And if you look at the cost side of it, from what I understand, the localization process is maybe 2% or 3% of the cost of production. So even if you take that back in-house and save almost all of the cost, it's not really going to change the overall cost of the production. It's not a huge driver of costs for these studios. So there's not a huge amount to be gained, but there's quite a bit to be lost if you get it wrong. So you need to get it right. And even if you were to reinsource it or take it back in-house, there's not a massive amount of cost savings to be achieved. Second thing is in terms of complexity. And as Stuart and Chris have described, there's a lot around this process that isn't just the translation, it isn't just the bits that can be easily pushed towards AI or machine translation. There are lots of technical processes, as Stuart was saying. There are lots of different parts of the works on the methodology that you couldn't easily replicate with AI, probably you could never be replicated in AI because there's a lot to them, and you really need that human interaction. And then the last bit is the creativity and making sure that you're trying to get that nuanced structure across. You're trying to make sure that you're getting the very best translation as the guys have described, there are lots of challenges and complexities there that AI will really struggle with at the edges. So I think there are those 3 particularly good reasons why the studios are perhaps less likely to take that work back in-house and why I think it will probably remain outsourced for the foreseeable future.

Okay. Just to sort of close the session, I was going to ask each of the panelists, what should people keep in mind about the future impact of AI on media localization and the broader entertainment industry? And how do you see ZOO as a beneficiary? I don't know Gareth, if you want to start?

Sure. I think these guys have described it very well. There are ways in which they can use AI within their business as I was just saying, I think it will remain outsourced. But as long as the customers see some of the benefit of that either in terms of improved processes, maybe better efficiency or higher levels of quality of work, then that outsourcing structure will remain in place and both ZOO and its customers can benefit.

And Chris?

Yes, I think we've got -- we've potential to again a huge benefit from AI. I think we have the potential to have a tool set that is very efficient, is -- it reduces the amount of errors that are within the projects. So I think for us, there's a huge benefit. And there's a potential that it could also open up new markets to us by having a more streamlined process.

And Stuart?

So just building on Chris' comments. I think as I said earlier, as technologists in our industry, as a leader in innovation in media localization, I think we're in a fabulous spot really to be able to identify where these technologies can help, and where they actually can be a hindrance because if you use it in a wrong way you end up getting a worse result out there than perhaps you have achieved before. But for us, the key thing here is to ensure that we're addressing what matters most to our customers, which, as I say, is all around quality and authenticity. So there can be no compromise for that for our target customers. And we can see there's also places where we can be using AI to make the job of our translators and our voice actors and so on more efficient, more rewarding to spot potential problems, call them at an early stage and so on and therefore, enable us to do what we do as a leader in our industry.

Thank you very much.

Good afternoon, everyone. Thank you very much for joining us for this webinar. As you may have guessed, we prerecorded the actual roundtable a couple of days ago just for practical reasons. But we're now joining you on a live stream for -- to go through a Q&A on the topic of AI in media localization. There are lots of people on the call, and you will find, to the right side of your screen and a tab mark questions, an opportunity for you to post any questions you would like us to answer. We've had a few questions pre-submitted. So we'll take those first. But at any point, feel free to jump in and send over a question, and we'll get to as many as we can in the 30 minutes or so that are remaining of this slot. So I'm joined by Chris. Welcome, Chris.

So first question that we have. Google Translate does a great job with my business letters and being understood when I go on holiday. Surely, it can be used for subtitling. Well, hopefully, that's the topic that we covered in the fireside chat. And I guess, in summary, the key points we're making here in our White paper, which will be published tomorrow is that, AI does play a role in media localization. But right now, its greatest value is in supporting traditional subtitling practices rather than in replacing them. And the reason of that is the media localization is far more complex than just written translation. So if you're using Google Translate to help you navigate around Paris or something like that, then clearly, you just need to be understood. Whereas what we're dealing with is dialogue that is spoken by folks who are creating an entertainment experience. And therefore, the words -- the actuals words they speak, are just one part of it. There is so much more involved and AI right now can't handle all those elements because it really just doesn't have any understanding, is the wrong word used here, but any access to those sort of -- those other dimensions. It is a holistic approach, and they're obviously a wealth of audio and visual cues that need to be taken into account, as I say, goes well beyond the words actually spoken things like, the tone which things have spoken, the sort of background of the speaker. So what kind of -- what kind of vocabulary would they be using? That all has to be authentic and true to the character and the plot and the storyline and the era and so on. It's got to be simplistic to the mood and the body language and all the kind of emotional impact that you get from a great acting performance. And to be true to the original, these are all the elements that have been taken on board and that form part of what media localization is all about. So systems like Google Translate and ChatGPT are very much text-focused systems. They don't have access to all those other subtleties that are needed to do a great job in media localization, particularly at the high end of the market, which is where ZOO is focused. So this AI, the current generation of AI, is more suited to literal written translations. And as we covered actually one of the questions that came up on the fireside chat, you can use machine translation as the first pass. But in our experience, it doesn't give you a great result. and AI can be used for things like creating captions in the same language for deaf and hard of hearing. But that's very different from localization. So it's important not to confuse captions, which normally mean sort of a verbatim same language text of what has been spoken with subtitles for other languages, which is -- and creating those is a creative process. So next question is, if you can use machine translation for a first pass, this must seriously reduce the cost of subtitling. So what's going to happen to ZOO's profit margins? Chris, do you want to take the first part of that and I'll maybe touch on the profit margin point?

Yes, sure. Just as kind of Stuart alluded to then, the first phase of translation is not necessarily just the verbatim, translating the text into a different language. Its actually taking and making it work in that particular language. So there may be particular idioms in there or book or comedy and jokes that are within that dialogue that need to be kind of localized into a particular language. And so what we found is actually by running translations to a machine translation as a first pass actually doesn't save you a huge amount because there's quite an awful lot of editing that is required after that process by a human. So we don't see a huge benefit right now of running machine translation as the first pass. It has got place and it's got there's elements where it can help and can speed up the role of the translator. But right now, it can't kind of -- it doesn't get anywhere close to what is needed as a dialogue localization kind of first pass. So we're not utilizing that right now in our processes.

Thanks, Chris. And actually, it might be just worth mentioning that there have been actually over the last few years, some really successful non-English original TV series that have been very popular on some global streaming services that received a lot of viewer criticism because of the quality of the subtitles. And it turned out of the fact that it was because AI have been used as a first pass in that process. So they just missed -- that approach just missed the cultural subtleties. They just weren't realistic. And in that situation, the translators have been paid less. So they're not being as attentive to actually the underlying meaning behind -- in the words. They're just taking those words on screen as suggested by machine translation and trying to maybe tweak them to perhaps enhance the grammar or whatever it may be, but that's not really what's at [indiscernible]. And in those particular circumstances, the result was that -- audience are turned off. And actually, the distributor had to redo the subtitle. So they actually did them again, but this time using trained professional experienced localizers who did a much better job. So on that margin question there. Obviously, technology enhancements generally over the ages have led to greater demand. And we think that's absolutely the case in media localization and is very much -- you're seeing it now where machine translation of subtitles and even automated dubbing is being used for user-generated content. So that's a whole new area that is opening up a whole new market segment that just would not have been viable before because someone who spends very little creating some user-generated video is unlikely to spend money with the professional localization company to produce subtitles. But if they can create them through a machine translation approach using AI, then why wouldn't you do that to expand your audience. So these technology enhancements like the AI type solutions that are coming into our industry. We believe that they will grow demand right across the industry. So we don't expect our margins to be materially damaged by these kinds of innovations. On the contrary, we expect volumes should grow as a result. So the next question I've got here is, videos of fake news are super convincing. Surely, you can fake it for dubbing with AI? Okay. Well, maybe I'll start that and perhaps hand it over to you, Chris. I guess the first thing is that, as I just explained, AI isn't really suitable as a complete solution for writing translated dubbing scripts for exactly the reasons that we've just covered when we've spoken about subtitling. I mean dubbing scripts and subtitles are different things. It's just worth being aware of that. They fulfill completely different purposes, and they're different and they can produce completely separately. So as a complete solution for sort of high-end entertainment, AI just isn't there yet. But having said that, there are a wide range of content types in entertainment. User-generated is obviously at one extreme, and we're more focused on the opposite extreme, which is very high end, very high value, very expensive to produce type content. But there's a host of things in between. And certainly, that there are some areas where absolutely these technologies can be used. So maybe, Chris, do you want to give a bit of insight into the areas where we see AI bringing value in dubbing?

Yes, absolutely. I think there's a huge benefit of utilizing AI as part of the whole process. And that's not necessarily creating synthetic voices, but it has a huge benefit for identifying certain things within the video that you want to look out for. Context and how that information what needs to be passed to the dubbing artist. And I alluded to it in the video that we put out as well is around things like child's -- children's voices, which are very difficult to record. It's quite problematic and quite a laborious process to get children's vices. So if we can utilize still having a human in the loop that provides the active element of it that can convert their voice into child's voice then it has a huge process benefit, cost benefit as well to do that. So we can -- we are seeing it come into a lot of the processes that we do. But as Stuart mentioned, it's not quite there yet in terms of getting the believable voices and the generative side of it and generating things from nothing. It's just not there yet to be able to replace that side of things.

Thanks a lot, Chris. Next question. AI means that the big studios can handle localization themselves. ZOO is going to be taken out the loop. I guess that's more of a statement than it is a question. Well, I guess the quick answer to that is the media localization, the vendors who provide those kinds of services, employ teams of specialists. And these are -- many of which are skilled creative. So the question is, is it actually financially viable for big studios to actually insource that thing? Is it practical? Given, of course, to cover localization means typically, a big budget title will go, will get adapted into sort of 40 to 50 different languages. And each one of those languages is in the main, non-overlapping with all the others. So that's to say, you've got to have a team of people who can cover French, [ Persian ] French. You got to have another team of people who can cover Canadian French, and so on for every one of those languages. And clearly, orchestrating something like that is quite challenging. So we don't believe that it's practical in the main for large studios to actually attempt to in-source that kind of capability. Given the quality and authenticity of the results is really the key thing that is needed for premium media and entertainment and the AI applications that are available today to deliver at that level always require a skilled human in the loop. And the question is where would you find someone with those skills and typically, it will be amongst those organizations who are, as part of their business or providing those kinds of services. Okay. Next question. AI is going to take over all the key localization jobs at ZOO. Another bold statement. So I don't know whether that -- is that one that you want to take, Chris?

Yes. I think it's important to understand what the roles are within localization. There's not -- we do employ a lot of people that are around -- that have language experience and can translate and dubbing artists, but there's also a lot of people within the roles to help manage the process to QC things to make sure it's -- the quality is at the right standard. And how we're seeing this is not necessarily going to replace these people's jobs, certainly it will have an effect on it, it may change their roles, but it hopefully will allow them to do more with the time and the skill set that they have. It's kind of similar to how we see it as part of the software development team we have. AI is not coming in to replace their roles and take that away but it allows them to give them the tools to do more. So we're looking at where we can exploit AI in our whole process from end to end really and where the best use of those tools are. We have a long rich history in developing automation and software tools for the industry. So we feel we're really well placed to be able to identify where those problem areas are where AI can be utilized to make people more efficient and very, very well work. We understand the localization process. We understand what our clients want. And so we're not blinded by the technology. We understand it's got its limitations and what those are and what our customers really expect. The standards that they expect and we were able to keep delivering to that standard. So we will utilize AI. We are developing tools internally that utilize a lot of AI technologies to be able to make it as better. But at this point, we don't see massively wholesale replacing localization jobs.

Thanks a lot, Chris. The next question is what percentage of ZOO's future business will revolve around AI? I think that's quite an interesting question. And obviously, it's impossible one for us to answer, but I guess we can offer up our thoughts. I guess I think, hopefully, one of the themes that's come across here is that as technologists, we are looking to deploy technology in any form really to the work that we do in such a way that actually it brings benefits. Benefits to ourselves, benefits to our customers, benefits to our collaborators like freelance translators, and voice actors and so on. So AI is a weapon in the armory, if you like, to go about doing that in the same way that automation has been a key part of the strategy that we've adopted from the very beginning in making the process that we work on more efficient. So Chris, I mean, I don't know whether you got on that question, what percentage of ZOO's future business will revolve around AI? Whether you've got any specific examples maybe things we've been doing that would be kind of bring that to life a bit?

Yes. So we've rolled out a few tools along our processes already, that utilize AI. Then one specifically, is an ASR system. It kind of automates the transcription process. So at the very start of our process, we get a video from a client. We have to transcribe that video into English, and then that goes off as an English template to be translated or utilized in dubbing. That's the very start of the process. It's -- nothing else can really happen until that piece is being done. And we've utilized AI to actually transcribe that video from scratch. And we've been able to save significant amount of time utilizing AI to transcribe things and then have a human pass over it to make sure it's accurate and the quality is at the level that we want. And that's significantly saved us time that would be hours and hours worth of work for individuals, but that can now be done in minutes utilizing the AI technologies that we can adopt in those areas. And that's where how we see we can adopt AI into other areas of our businesses where we can do more with the resource that we have available. So one example of something that we put in place already.

Thank a lot, Chris. Next question. Who do you see as the leading developers of AI localization? Small niche specialists? Or is it the big media AI giants? And who should we be keeping an eye on over the next few years? I'm sure, Chris, you've got quite a few insights on that. Maybe I'll just begin by just, I guess, clarifying something if it's not already obvious that there are a small number of very big companies who are busy building platforms and investing in huge amounts of hardware and infrastructure and building these models that are put together by sourcing enormous quantities of original, what's referred as training data. That essentially the information that's out there already from which the AI software can infer patterns and use those to produce new things. So there are a handful of big players who are doing that thing. That, if you like, almost building a sort of a generic capability like OpenAI and Amazon and Meta and Microsoft and so on. And then what you see actually all around the world is thousands and thousands of start-up companies in many cases, who are basically building vertical market applications on top of those general platforms. So these are organizations who have domain expertise in a particular area and have knowledge and insight and quite possibly have sources of data that can be used to augment or adapt the -- these large language models or whatever they may be, so the -- in order to tailor the outputs to their particular domain and area of specialism. So Chris, I don't know, obviously, we've been actively following this for 10-plus years now, and Chris and the team have been evaluating various technologies along the way. I don't know whether there are any that you wanted to highlight, Chris?

Yes. I think I should say, we're kind of probably be on the realms of small companies developing their own models now. They've got -- they're in the realms of millions and millions of dollars invested to be able to develop new models. But actually, the -- as you said, a small number of companies who are able to take these technologies and develop them and can refine them into tools that work in a particular sector. And we feel we're one of those, as you said, we've been in this for over 10 years now, and quite heavily over the last 5 years where we've been developing our own technology and utilizing the debt that we have to be able to refine and build applications, build models, refined models with -- that's very specific for our application really. So I think you're going to find that there are small specialists in this area, but quite a few are in the same vein in terms of what we are as a business and that we are a service -- we have a service offering as well. And we've been in the industry for 20 years. So we have a good insight into how these things work as well as technologists as well to be able to apply that in action really. So we're quite well placed, I think, to be a key player in AI localization moving forward.

Thanks, Chris. And the next question is kind of touches on a similar topic on competition. So the question, two parts to it. Do you think you're ahead of your key end-to-end peers in using AI? And how long have you been researching and developing AI applications? Is that something you want to take, Chris?

Yes, I can do that. As I mentioned, we've been working in collaboration with academia and a number of leading companies in terms of AI for 10-plus years now, it's something we've been developing for a long time. So we do have a good knowledge of what's going on in the industry and how we can utilize that. We built a team over the number of years of software engineers who specifically experienced in AI technologies as well as data scientists to be able to gather and utilize the data that we have. So we're quite experienced in that area. So I think we are kind of ahead of the curve in terms of what we're doing. And that's always been ZOO's approach. We always look at things from a technology point of view. We don't rest on our laurels and just go deal with how the industries go. We'll look at how we can disrupt. We look at how we can develop technologies that will kind of add to what the industry has already. So I think we're in quite a good position compared to a lot of our competitors really.

Thanks, Chris. The next one is 2 parts, which I'll take one at a time. The first part is what are clients expecting from you with regard to potential savings that could be passed on? I guess we haven't really spoken about savings from, I guess, a pricing standpoint in terms of direct benefit to customers so far in the Q&A. But obviously, customers are hearing about these technologies and anticipating that a greater level of automation may be possible in the work that we and our peers do that could yield lower prices that obviously. Although as you said, our customers value most of all the kind of quality and authenticity of this work. Of course, anyone is very happy and pleased to find savings whenever they can. So I think that -- I think, generally speaking, our customers are keen to explore, can these technologies be brought to bear in such a way that they would benefit in terms of lower prices, lower cost to do this kind of work. And certainly, with several of our customers, we have been in dialogue and doing tests and exploratory type things to help them to understand the scope, the capabilities, the limitations of the current technologies and where they could be used. Which kind of dovetails into the second part of this question, which is, are clients realistic about the savings and the potential applications of AI to premium content? I think when -- my impression is that when this sort of new wave of AI technologies appeared 1.5 years ago or so, I think it's fair to say that in the entertainment, we do in the same industry, lots of big executives and big companies certainly took interest in that and imagine that this could be transformational for them and enable saving a lot of costs. What we find today, 18 months down the line, now that these organizations have seen what the software can do and what it can't do where its strengths are and its limitations. In our end of the market, in that high value, high end of the market, the conclusion is what we hear from all of them is that AI isn't there yet. and maybe it will be 1 day. But right now, they're not prepared to take the risk of instructing their vendors to provide a different service, a lower-priced service on the understanding that AI is used predominantly for -- to fulfill that service because they are very concerned, and for good reason, the result would be in theory to the way in which it's done now, and it would be a shot in the foot because at the end of the day, the aim of what we're doing for our customers is to extend the reach of their products to bigger audiences. And if certain audiences are alienated because the way that they experience that content through the localized materials is inferior, then it will turn audiences away and that will be counter productive. Next question on a similar vein. What are you hearing from your customers about their acceptance of AI within the localization process, which I think we probably touched on. But the second part of this is presumably ownership of the intellectual property rights might be a stumbling block? We've already seen this begin to play out with OpenAI and Scarlett Johansson. Yes, for those who haven't followed that particular storyline, here was a case where a company used the -- not an active voice of Scarlett Johansson or not Scarlett Johansson's own voice, but actually apparently, someone who is impersonating Scarlett Johansson to basically provide voice over and such things. And they subsequently withdrew it after a -- this was flagged up as an intellectual property issue because even though it wasn't Scarlett Johansson's voice, Scarlett Johansson is an actor who makes her living out of her lightness and her voice. And therefore, the use of AI in that way was depriving her of legitimate income. And -- and that's -- it's not a terribly well-chartered area, but clearly, there is a lot of attention on the legal implications of the use of AI, given the way in which they are trained. So as we've touched on already, these systems are trained using pre-existing information, some of which may be copyrighted. So the question is, if you use an AI system to produce something new, but that new thing, of course, has got elements of preexisting things in it, then where does the intellectual property right in that output set? And also, is it -- is there -- has there been an infringement of intellectual property rights in its creation? So it's -- that is a whole legal minefield that's playing out at the moment. And certainly for our customers, what we hear from our customers is that they are explicitly requiring us to be upfront with them and to only use AI in situations where they -- that they understand and that they give their consent. And by -- to give their consent, they will need to understand the legal implications of that and ensure that they are -- they could not be subject to a legal corporate infringement claim downstream. We go to the last question, and our time is almost up. So hopefully, we've got time to cover this. I was just wondering what your latest thinking is in terms of how AI might affect your economics? If your cost is potentially reduced, do your customers still focus on the value you're offering and hence, potentially might it provide scope for greater profitability? I think Chris has spoken and given some examples of some technologies AI, technologies we're already using that are giving us efficiencies. What our customers care about the most of which we said already is the sort of quality and authenticity of the output, first and foremost. But secondly, it's about time to market. They want to -- what we're hearing from all our customers is that they want us to work faster. They wanted to get the results of our labors sooner so that they can then exploit their materials internationally. So normally, in this industry, if -- to do that, if a customer comes along and says, I want this in half the time that the SLA requires, then you would -- they would -- the vendor would charge rush fees. So there'd be a surcharge to pay in that situation. And the client, if they want to get these assets very quickly, would just accept the fact they have to pay extra to get that. But we think what AI is doing here is potentially reducing the time it takes to do this work in such a way that potentially clients can benefit from getting the results more quickly without having to pay those surcharges. So there is -- I think there is a sort of an economic point here that is a great value to customers because of that real drive to get product ready for market more quickly. So with that kind of wraps up all the questions we have. Thank you very much, everyone, for attending and for submitting all the questions. I hope we've covered everything satisfactory. This recording will be available shortly online if you want to refer to it again. So with that, I'll say goodbye. Thanks very much.

Thank you.