14 | Create Engaging Videos for Marketing and Customer Service in Seconds, With Yuval Machlin, an AI product expert

0:00Hello and welcome to Leveraging AI. This is Isar Meytis, your host. And in today's episode, we're talking to Yuval Mahlin, who's an AI product expert. And in this episode, we're going to talk about how can you create a quote unquote talking face video in seconds instead of days or weeks. And what specific use cases is this perfect for and how to implement it for these use cases. At the end of the episode, I'm going to give a quick summary of interesting news from the AI world that happened this week. So stay tuned for that. And now to the episode with Yuval Mahlin, where we dive into video creation in seconds.

0:39In the next few years, AI technology will change our world dramatically. Whether you are a business executive trying to catapult your business forward, or just somebody who refuses to be left behind and want to advance your career, this is the show for you. I'm your host, Isar Maitis, a serial entrepreneur and an AI enthusiast. You'll hear invaluable practical tips from innovative business leaders, AI practitioners, and some of the brightest AI minds in our world today on how you can leverage AI in ethical ways to advance your career and grow your business.

1:20Hello and welcome to Leveraging AI, the podcast that shares practical ethical ways to leverage AI to improve efficiency and drive business growth. This is Isar Maitis, your host, and I am truly excited today. And the reason I'm so excited is because video is a very big part and component of any communication channel today, whether it's personal and business. And since we're talking about business, it's whether it's internal business like employee training or external such as marketing or customer service, et cetera. So video is a big component and it's going to be even a bigger component moving forward.

1:56The problem is creating videos has a lot of necessary evils that comes with it, such as a lot of logistics, like you need a studio and actors and a script and videographers and setting up lighting and microphones and so on. Editing takes a lot of time. The review process is very far from efficient because you have to have at least a good draft to show it to decision makers. And if they make major differences, it's not just back to the editing room, it's back to the studio with the logistics we talked about before. If you're running an international company and you want it in multiple languages, well, this multiplies itself in the amount of languages you want to run this at, including different reviewers who are not necessarily people from the company because those people may not speak the languages you want to use.

2:45And once something changes, whether it's your product, the market, the industry, anything, you need to go back to the beginning and do this thing all over again. So all of this goes away if using AI-generated videos. So you gain all the benefits of video communication, the personal connection, the delivery method that is more personal, but without all the overhead and problems that come with traditional video creation and editing. And since this trend is just going to grow, meaning there's going to be more and more video across everything that we do. And since generative AI is now allowing us to create those videos basically on the fly in seconds, I am so, so excited for our guest today.

3:34So our guest today is Yuval Mahlin. He is an entrepreneur for years. He has been leading product development and strategy of new products for many, many years. In his recent company, he's leading a development of a product that allows to create videos from scratch. And we're going to talk a lot about this during the episode, but it's really nothing short of magical in my eyes what this product can do. He is considered a global expert in AI-generated videos because of his role in leading one of the leading companies in that field. And hence, I'm so excited and humbled to have him as a guest on the show.

4:10Yuval, welcome to Leveraging AI. Thank you so much for welcoming me. So before we dive into the use cases, we have three unique use cases. You know, the first one is internal communication and training, the second in marketing, and the third one, which is really exciting to me. And I'm actually surprised with myself that I'm excited about customer service, but it's real-time customer service. But before we dive into that, for the people who hasn't played with these tools yet, what is AI-generated videos? How does it use? What's the interface for people who hasn't seen this before? So first of all, it is a very, very, very exciting subject.

4:51And I think that, you know, we are at this tipping point when it is actually available for anyone. And as you mentioned before, you know, this process is so tedious to prepare videos and to go to the studio and to select the actor or actress and, you know, and so on and so forth. And the generative AI just make it so simple for people. You just put an image of a person and add a text that you would like them to say, any language, any accent you can select, and any flavor that you want to add to it. And you just click a button, and a few seconds afterwards, you have this video of the person speaking the text.

5:35So how it is made? The AI is driven by data. So the model learns based on many, many pictures and videos of people speaking any language. And then it generalizes this behavior. And then it can, based on this learning, it can produce the required video that you want. So to paint this picture for people, you can take either a pre-existing avatar that looks very realistic. And the benefit of that is it can come in any ethnicity, age group, sex, clothing, background, like whatever setup you want, there's a huge variety to pick from. or like Yifal said, you can take your picture or the picture of anybody in the company or outside of the company that you have the permission to and use them.

6:28And the AI knows how to change face expressions, leap movement, eye contact, head nods, anything that a human would do when speaking on specific subjects. And you can peek, like Yifal said, the language and the level of excitement in the conversation. and add a script to it and it will produce the video in literally seconds. So as I said before, this is the closest thing I've ever seen to magic. So let's talk about the specific use cases. How do companies today, like you said, this is not the future. This is stuff that's happening and available right now. How are companies that are customers of yours or other companies that you know of are using this today?

7:15Yeah. So let's start with the challenge. So at any organization, we have training materials, we have presentations and documents. And the reason that we prepare those is because we want to transfer the knowledge for our colleagues or customers or anyone at scale. So you write once this presentation and then you send it to whoever you want to send it. The challenge there is that presentations are either too short or too long. So, you know, we learned a lot about making it clear, three to five bullets each slide, because our mind cannot, our brain cannot capture more than that. And then you find yourself with slides that only the presenter understand.

7:58Or you'd like to be more precise. So you add more and more explanations and text, and then you find yourself with something that is really, really hard to read, and no one really reads it. So adding to that, the presenter that reads, not only reads, but tells you about what's there, then you have the knowledge transferring. So people did it before AI. And the way they did it, they went to a studio, they brought a cameraman, you know, and everything else. And it's really, really, really expensive and it takes a lot of time. and the bad news, even if you have the money and the time, that it is obsolete the second you finish because then you want to change something, you need to do it again, right?

8:43So AI solves everything in that manner. So you don't need to dress nicely. You can bring your own photo and change the clothes as you wish. You can even generate the presenter themselves by using a text, a very specific text, and then you get the presenter. or the AI generates the presenter for you. And then the text, so that AI can help you generating the text, but you might want to do it yourself or maybe you have previous materials you can use. So these are things that our clients are doing right now. And the way to do it is just, you go to the studio, we have a self-serve studio, you upload the photo or generate it.

9:23Again, just to clarify, it's not really a studio, it's a web application. It's a web, yeah. We call it self-serve studio, but of course it's a web app accessible via yeah of course any browser because like oh we're going back to the studio no it's just a web application yeah it's a virtual studio actually it's not it doesn't look like a studio of course it's just a web app upload your photo on the photo that you want to play and then generate the text or upload the text click a button you get a video then you You can download it and put it on your preferred presentation tool, whichever you like.

9:59And that's it. So again, I just want to highlight how incredible this process is compared to what used to happen in the past. It goes from weeks to seconds. That's how dramatic the difference is. So between all the stuff we talked about before, which I'm not going to repeat, it goes to I need the text that the presenter needs to say, and I can choose a virtual presenter or upload my own image or somebody's image, click go, and the video exists. And going back to creating the presentation, and now let's go back to the actual use cases. You talked about training materials. So you can today take, and I'll let you dive more into this, but you can today take a lot of your training materials, upload them to ChatGPT, ask it to summarize it as a presentation and load that to your quote unquote studio.

10:59And that's it. And you're basically done. You have a presenter speaking about the stuff that he needs to be speaking about. Can you give us a little more color on how companies are using this today that you have seen from actual implementations? Sure. So first of all, it's even simpler than that. So, you know, you have your materials, the company that has the materials. And now they want to use the engagement with the materials and they want to increase the level of knowledge in the organization. So having this presentation already and you have the summary because you have usually someone that is responsible to transfer the knowledge, they write the summary and then upload the photo, click button, you get this video.

11:45So they embed it into the presentations, and now they use the same, the old and good materials that they trust with an additional layer that it's like, you know, we would like to be one-on-one if we could with every one of our employees or customers. And now, I mean, this is the way that they do it. So we see it all over. This is the main use case that customers are using right now. And the reports we're getting back is that the employees understand much better the materials. and this, I mean, I think it's very intuitive, but even when you come back to an employee a month after, they remember it much better than before.

12:27Yeah, so knowledge retention. Knowledge retention becomes... Yeah, exactly. You have the human aspect and our brain is prepared and react much better with smiling face, not to say talking face. Interesting. So I think what I really like about what you're saying, and in this podcast, this is going to be a repeating thing, I'm sure. The main thing when you try to implement a new thing into an organization is not to try and create a revolution. Try to stick to your existing processes, existing procedures, existing tools as much as you can, because then the adoption is a lot easier and you don't have to go and teach everybody how to use this new thing.

13:11they use the same things they've used before the same presentation in the same content management system in the same learning management system with the same tests in the end but now you're just enhancing it with an actual speaking person that instead that you need to generate it in like you said weeks you're generating it in seconds and you're getting like you're saying validated better results both in the short term and in the long term which is incredible exactly adding to that also one one more thing which i think is really important that i saw in some areas that really provide good results the viewer they can change they can define who they want to learn from oh awesome so so basically you're saying you have this video produced by a variety of these avatars and the student can pick the avatar that he wants to be his quote-unquote instructor yeah and the student can even define by themselves almost in real time even if the video is not prepared ahead of time because you have the text and then the student can say hey you know why not learning from you know this and that person as you said it's a magic and it's this incredible like i now i'm still thinking old school right i was thinking well you got to record five different variations of the video, but he's like, no, he literally created a video as the student is selecting a specific avatar.

14:44And amazing. Amazing. Okay. Let's talk. So this is for corporate training. I think it's very, very clear what are the benefits. And we talked about another thing that is very important. Corporate training changes. It changes because it could be, you know, there's a new product. So you need to teach your customer service people how the new product works. It could be because regulation change and now you cannot say this, or you have to say that. All these things, instead of going back to a real studio, you literally just upload a new text and you're done and you have your new training ready to go.

15:17So I think that's another huge benefit from a training perspective. Let's talk marketing. How are companies using this from a marketing perspective? So in one of my previous roles, actually, I was a VP of product at a company that the main expertise was marketing. Actually, it was Albert Einstein's exchange. And there, we always talked about how text can be hyper-personalized. You have today the marketing systems that they know which customer is coming. They know the segmentation. And in many cases, they even know your name because you're part of the ecosystem. And then they send you messages, text messages, with the exact text that they believe is relevant for you.

16:02Now, video, on the other hand, is static. You need to prepare it before, but people react much better to videos because it's fun and it's engaging and it's with the visual aspect. So these days, the first time in history, you can actually combine the two. So you can create hyper-personalized videos using the hyper-personalized text that you already have with the video, injected into the video with the very specific presental that the company believes that is best for you. So, yeah. This is mind-blowing. So again, I'm going back to the holy grail of marketing, right? The holy grail of marketing is to have the most convincing conversation you can have with prospects and clients and so on.

16:54and it's just not possible to do this, right? The solution for this is to have an army of salespeople who actually do this, right? So you want to grow, you add more salespeople who can have one-on-one conversations and explain to that particular individual that comes from a mid-sized company in the Midwest US that has two employees in his company and that has this amount of budget, it will be a very different conversation than somebody who is running a division in a corporate in Japan that want to use the same tool. And obviously, so far, it was literally impossible to do other than to have an actual person.

17:34And what you're telling me today, based on digital data and footprints that we have today, we've been using to build funnels, we've been using to build direct email, we've been using to A, B test stuff on the website, personalization, and so on. You can take the existing data as far as the personalization needs, text, messages, and so on. And in quote unquote, real time, generate a new video that will be, if I'll take my two examples, a person from the Midwest talking about the relevant stuff to a mid-sized company with two employees. And in the same exact website, without really changing anything, the presenter is going to be Japanese speaking Japanese to the head of division talking about division relevant stuff.

18:17It's literally incredible. Yeah. So I want to ask a question because we're in the very, very early stages of this. You're in this because you're in a company that provides this as a service. How many people are actually doing this already? Not thinking about this, but implementing the stuff that we're talking about right now? So without exposing, you know, real numbers here, but the trend is, the trend is going very, very nice. And it's still, it's still the early days. So with corporate training, I think we reach to, you know, a lot of our audience, many companies are doing that already with the marketing, it's still in the early, but there is a very, very big interest.

19:01And, you know, you will see it around us. We will see it around us very, very soon. Yeah, I think the interesting thing would be that anybody who will understand that adopting this would lead to A, lower costs and B, higher sales. So you win on both ends of the equation, will implement it as quickly as they can, which by definition means smaller organizations will move faster, which is going to be really interesting. I think we're heading into a very interesting time. I want to talk about...

19:59to get personalized messages than, you know, generic, but some people think differently. No, I agree with you a hundred percent. I think, you know, a lot of, it's the same conversation every time, just with different technology, right? It's like, do I allow cookies on my browser or not? Exactly. Oh, they're going to collect data on me. Well, they're going to collect data and then I'm going to see ads for drones and sport equipment versus for feminine products that I'm probably not going to purchase. So there are benefits to going forward with allowing those platforms to work for you because that's what they're trying to do.

20:36They're trying to work for you. They're trying to make your experience a better experience. Speaking about a better experience, let's talk customer service because the use case you showed me over there literally blew my mind. And I'm in this field and I'm reading a lot and I'm watching a lot. This is incredible. So talk to me about what's going on in customer service right now with live video? Yeah, so the real-time for a customer experience needs to be really real-time. You know, it's not seconds, it's milliseconds. And we're almost there. We are almost there. And the idea and what we now see is that you can generate not only the presenter and not only the text, but to combine the two in real-time.

21:20So when you've got a website and you want to have customer support or you want to have some help with a representative. And the only way to do it at scale is when you have the AI learning the material and then understands your questions in real time and then reply to you in real time using this amazing technology. And you also see the preferred face. So this takes the customer's support and experience to, you know, new frontier. So I want to touch on three different aspects here that are, again, everything we're talking about is a complete revolution. It changes everything we know. But in this particular use case, if you think about customer service, the two most frustrating things that happens to us as business people and as personal lives is customer service.

22:17Why? Because you always have to wait, always. It's never instant. Even if it's chat on the website, it's always, you're always chatting with some kind of a bot in the beginning, and then it transfers you to an actual person, and then you have to wait for the person. And that person actually supports 10 people, so he's not really with you the whole time. So his responses take a little bit of time. So that's frustration number one, it's just time. Frustration number two is in some cases that person doesn't know. And he's like, okay, I'll get back to you via email tomorrow. or let me ask my supervisor or let me transfer you to the finance department because it's a billing question and not a technical question.

22:53And it drives you absolutely crazy. And now what you're telling me is that there's this virtual entity that literally knows everything the organization knows about the potential of customer service and is there all the time because it's a digital entity. There is no limit to how many of these you can have. So the outcome is an immediate respond with the perfect answer right as you click on, I want support, which going back to giving up on privacy and the trade-off, if you give it access to your file, to your history, to whatever the case may be, it also knows everything about you. It knows who you are.

23:36It knows what services you have. It knows how much you're paying. It knows when was the last time you paid. It knows what products you're actually using. It literally knows everything. So its ability to actually support you is significantly better than an actual person on the other end. But then you showed me something that really blew my mind, which is it's actually active. So talk to me about this. So first of all, you know, it's step by step. You know, the future that you're describing, yes, this is what we aspire, this is what we want to have. It will take faster than most of us believe, but it's not there yet.

24:14Okay, so it's an assistive tool. It's much better than an average person, at least what I experienced. But still, you know, it's an assistive tool. It solves most of the challenges that you mentioned. Like, you know, you don't need to wait anymore and it feels realistic. and it has access to the data that you already provided to the company that sells you because this, and it's not related to privacy at all because this is the data that the company already has in order to support you. There are some cases that you need to gather information from many, many different places. The AI does it very, very well, but it still supports the real people that provide the service.

24:54So it's not yet independent. It can also be proactive, yes. So let's say that you come to a service and you'd like to achieve a goal. So you can speak with natural language, you know, using your voice and ask the question. And then the service analyze the exact situation that you are at the same moment and can offer you like a solution that will not interfere with your current experience. I think that the example that you mentioned is like the video that they send you about those people that are watching TV, right? And they kind of, they need help to get something that was wrong in the TV during the show.

25:38And then the virtual assistant solves the problem, but it requires payment. So they solved the problem for this very moment. And then they say, okay, I'll reach out back in the halftime and then we'll finish the payment. Yeah, this is an amazing experience that I think we all would like to have. Yeah. So again, to clarify that video, what it shows, it shows a bunch of guys watching football and the internet speed is not fast enough and it's jumpy and they can't watch the game and they turn on customer service. It turns on in the TV as an overlay on top of the game from the ISP, right? So it's the internet service provider jumping like a virtual person saying, oh, I see you having issues.

26:19I can upgrade your speed. Would you like to do that? He says, yes, I'll upgrade you right now. Watch the game. We'll deal with the payment afterwards. And you think about it and like all these capabilities are here right now. I don't think anybody actually implemented it yet, but the every part of this incredible functionality, and you can apply this to anything, you know, to try to get a doctor's appointment, to changing your speed at the things, to buying tickets to a show that you want to watch and trying to figure out what are the best seats and what's available for that week versus next. All of this is literally going to change dramatically because there's going to be a person at the other side that knows everything, that can take all the relevant actions, and that is available 24-7, 365 days a year because it's a virtual entity.

27:15So I think this is all incredibly exciting. I think we're going to be seeing dramatic, dramatic changes in the use of videos. If people want to do these things today, if they want to play with it, what tools are available right now that people can go to on the browser and start playing with these kind of tools? So in terms of the turning text into video, turning a presenter into a speaking video, go to d-id.com. We have a free trial over there and, you know, it's accessible to anyone from embedding or create generating the text. So I use the GPT. Use Bing, by the way, who implemented GPT into it.

27:58Very, very simple to use. I enjoy talking with this tool. Amazing experience, really. And also you have Me Journey. Me Journey to generate images, text to image. Very good experience. And you have many others, but these are the three that I really, really like. So again, to combine the three together, so for people to understand, you go and create text with ChatGPT. You can create a face that doesn't exist with Me Journey, and you can define what characteristics that face is going to have. You upload these two things to the ID, and you have a video of that new face that you've created that can have whatever features, whatever skin, whatever age, whatever, anything you want, speak the text that ChatGPT just created for you on a topic that you wanted to create.

28:48Another one that I mentioned just because it's exciting news that got out this week. And by the time this goes out, it will be old news because things are going like crazy. But Runway, which is an amazing, amazing company that has a literally crazy set of tools for AI visual stuff, whether it's creation of images or manipulation of images and videos and so on, just announced that they now have a text to video capability. So just like you create images in MidJourney, you can create videos with no reference. So, so far, the tool still did magical stuff. Like you could take a static image of a person standing under a tree and say, I want that person to walk under the tree and it will create a video and the leaves would move and and the clouds in the sky would look realistic and so on.

29:36But now they're saying you don't even need that. Just like any other prompt, be very detailed with your prompt and it will create the video for you, whether it's a realistic, cartoonish, whatever you want, which again, this is where everything is going. So I don't know what videographers will do in a few years or video editors will do in a few years or camera manufacturers will do in a few years, but I think they will do a lot less they're doing right now. And they'll probably find other ways to make a living. I assume that's going to be very similar to a lot of other professions that we know today.