Transcript: Lior Sassy | Lightricks | How Do You Build an AI Product for Video Production?

Eyal David: [00:00] Hi friends, this is Eyal. You've reached Product Builder, where you can listen to the most interesting product conversations. My journey began in 2012, when I learned what product is at my own startup. Since then I've worked in the industry, and since 2017 my company has been advising a range of companies. My drive is to help all of us get better as product people. And before we start, if you enjoy listening to Product Builder, please rate us five stars on Spotify. It gives me great feedback that the podcast brings you value, and it would make me very happy. So let's get started. Hi Lior, how's it going? Good, how are you, Eyal? Great. With us today is Lior from Lightricks, PM lead in the Image vertical. Let's pin that, and in a moment we'll understand what the image vertical is. Deal.

Lior, LTX Studio and hiring juniors

Lior Sassy: [00:46] So tell us, what do you actually do at Lightricks today? So I'm, as you said, a PM lead. I'm on the LTX Studio team, a Lightricks product that launched in the past year, even less; I think we're now nine months into operation. And I'm a PM there together with two others, and three more are joining the team soon, very exciting. Wait, before we go on, you're also hiring juniors, right? Right. We, so personally I came to Lightricks not from experience; I grew through what's called internal mobility, and Lightricks really believes in that: taking people who are exceptional in some field, data, design, who are curious about product, interested in it, with analytical thinking, and bringing them in and building them up with us. So this is really the first time we're opening junior roles to outside candidates.

Eyal David: [01:38] Gorgeous. So we'll come back to that at the end, leave details and everything. Cool. So actually Alon was here in one of the first episodes,

What LTX Studio does: AI storyboards for filmmakers and ad creators

Lior Sassy: [01:45] also from Lightricks, but what you do is completely different. Right. So tell us a bit about the product. I'll tell you. So at Lightricks we're divided into several verticals that focus on different types of users. Alon, who came from Facetune, that's more consumers: people on mobile, people who are the business themselves or have a private business, as they say. And LTX is aimed at people who have a very specific use case, and that use case is making a film or an ad. These people are usually advertising agencies or filmmakers, whether small or large. And we try to fit into their entire process. So you can look at the stage where I just have an idea and I'm debating whether it's a cool idea or not, and also the stage where I actually need to show it to investors: I want money for this thing, I want to produce images and video for this thing. And we come in across the whole process. So basically what happens today, both of them, filmmakers today, or until today, would sit and sketch the storyboard in a notebook. Right, absolutely, I know it from my wife, who's a photographer. And now we manage that whole thing: with AI we actually generate that grid of squares, or whatever it is, of the different scenes, with actual actors, all of it from a prompt. Exactly, basically text-to-image is the thing. Text-to-image is me telling you, make me 1, 2, 3, and you make it for me. And we realized there's this market of: I say what I want to someone who's my hands and feet, and they draw it for me. And both with that storyboard artist and with the person initiating the project, no matter which of them uses it, it will save some

Eyal David: [03:29] some friction with the thing. And basically you identified a market here because, what, because during COVID there was a rise in content creation, and then basically, how did you actually identify it? How did you identify it? Good question. The truth is that, first of all, we were the market.

How it was born: DALL-E and a three-person internal tool

Lior Sassy: [03:46] Let me take you a step back to how it even happened. We got access to DALL-E, the first text-to-image there was, I don't know who here knew it and knows it, but our CEO had connections and got access to this thing, and understood, OK, there's something here that's going to be a very significant revolution. Also, we're a company that works with pixels, that's our passion. But you started from a solution, that's what you're saying. We started from a solution, absolutely. So we saw this thing, we saw text-to-image, we understood there's an upheaval of the image world here. And the LTX team, what's called LTX today, was a very small team of people trying to understand, OK, what's the potential, what is this thing, what can it do. We had one marketing person, our CEO, and one developer, and they simply created a platform meant for internal use, to produce storyboarding and a shot list before shooting our ads, before interviews for ads. Basically our marketing person was also the one bringing the need, in a way. They were really excited about it. They called us into a room, actually took us out of the roles we were in, told us they're founding something new, took us into a room. I was like, what's going on here, why am I starting a new product now that I have no idea what it is and nobody has checked or anything. They showed us this platform, which is really like a storyboard, just digital, and we were stunned, because the process there is: I can write a sentence, the smallest sentence. I can write, give me a romantic comedy in the style of Wes Anderson meets Emily in Paris. And something comes out, and it's really exciting, even as a personal experience it's really exciting. And then the question of how it goes to market came pretty fast, because in the end it served us too.

Eyal David: [05:35] So basically it started from your own need, and it also clicked for you because you probably knew it from the past. Do you usually have some kind of innovation unit in the company that goes and samples things, and then when something matures they call everyone into the room and say, let's understand for a second how to use this?

Lior Sassy: [05:50] Absolutely. I mean, we've always had research people, and Lightricks is a company that relies heavily on the brains of its people. We have super strong people, and that's part of our competitive edge, I'd say. Once AI came into the market so strongly, I mean, Lightricks has pretty much always worked with AI, even when it wasn't cool and nobody talked about it, but it really hit the market hard. And today we have, for example, channels called AI News where people just share every new thing and send summaries. We're constantly, constantly trying to be as up to date as possible. So absolutely, innovation is the name of the game, it's part of the thing. And you're active in AI News, how does it work? I'm active, yes, sure. You know, you're on Twitter, X, scrolling, you see something cool, you send it, people say, I'm working on something similar. You basically create a kind of community, and I think the whole company is there today.

Eyal David: [06:48] Right, I wanted to know what your information sources are, so OK, you answered me. So it's worth following you on Twitter too, we'll leave a link afterwards if there is one. So describe for me for a moment: you're in the room, there's the big excitement, applause, everyone says yes, this is what we want, and then what?

From POC to product: rebuilding from scratch

Lior Sassy: [07:05] And then we started working. That meeting in the room wasn't like, come on, let's ship a product. It was, we have an idea, we want to know what you think. And I remember that I and the product manager with me, Adi, sat there and we were like, wow, come on, yes, let's start making money, people are going to love this. But it wasn't like that, because that's not how it is with products. You know, you see something amazing and then the developer says, no, it's a POC, it's not there yet, ta-ta-ta. So we basically rebuilt it from scratch and added lots and lots more capabilities. It was also really a sampler of everything Lightricks can do: we know how to do face swap, come on; we know how to do lip sync, come on; we know how to do sounds, let's add it. We started to understand all the things needed to produce material, which for these people is a storyboard; there's a thing called a sizzle reel, which is like an imagined trailer, like

Eyal David: [08:05] this is how I show you the vibe, the genre. So basically we tried to understand what's needed there. That's crazy. You basically took capabilities and maximized them, absolutely. Because even when you talk to me about, what did you call it, a sizzle reel, even if I can produce a really good video that's short, but if it's long... So you took lots of things and maximized them into some product promise whose market wasn't yet clear, or you hadn't done research yet, and now you need to take it and actually build the case and start narrowing down. So that's the beginning of the work you did. Absolutely, both understanding what the case is

Beta, partners and the competitors: Runway, Midjourney and Krea

Lior Sassy: [08:41] and understanding what the capabilities are, which was tricky, because we saw it, and it's really easy to get excited and say, OK, you can make films, we've arrived, come on, let's start making Hollywood films with AI. But it's not there, and that's fine. By the way, it's not that it's not there; there are simply AI films, it's a genre today. Today it's a genre in its own right. I don't know if at some point it'll be a genre... I was at a film festival in Amsterdam, it's insane how much of a message you can deliver with this. But again, I'm coming back: for our people, filmmakers, ad creators, it's not there yet, and we also needed to understand where we can cater to them, where we can give them that value. And that was one of the challenging things: lots of conversations with partners, people who, in exchange for using the platform for free, gave us a lot of their knowledge. That was insanely cool. So how long have you been working on this product since that day? Something like nine months. It's still in beta, which is also something new for us at Lightricks, I can say, because we used to really try, we wanted to bring things polished and release them when they look wow and work wow. And because of AI, and because it's so community-driven and everyone gives and shows and shares their knowledge, we said, come on, we also want to enter this world and give the not-so-polished version so people can use it. And maybe there's another argument, and you'll fill it in for us: who are your competitors? Wow, so there are a few, and they're not exactly direct competitors. First of all, I'll say their names, it's not a secret: Runway, Midjourney, Krea. They're the big companies in text-to-image today, and they are, in a certain way, our competitors. I will say that today they're a bit less in the worlds of storyline and consistency. I mean, text-to-image or text-to-video for them is a single asset, and we see it as something more continuous. And that was also our promise: characters you can use throughout the project, and we're also thinking about it across several projects, a very specific and consistent style. So those are our strong points today.

Eyal David: [10:59] So I don't know, when I think about it, I expected you to say, OK, there are all kinds of small competitors that are very vertical, and Adobe, and Adobe bought someone who does this. So it's not like that, you're saying. No, actually, somehow people came in and became, like, companies that became giants in the market that you'd never heard of before. Krea, I didn't know about them, today one of our biggest competitors in the text-to-image world.

Lior Sassy: [11:23] But it also gives us lots and lots of hope about where our place can be. Because it's a completely new market and the space there is wide open.

Eyal David: [11:33] That's what I was trying to say: probably one of the urgencies there, to start fast, was, let's be the first to jump in the water. It was cute, absolutely, right. So tell us a bit, we talked about the industry; the industry probably also improved a lot in those nine months, right? Things happened, I see it in other tools too. What does your day-to-day actually look like, how do you manage to keep up and stay on top of all this? So it's very hard.

Keeping up with AI, and from theater to UX to product

Lior Sassy: [11:58] And part of my day really is looking, like I told you, scrolling Twitter is part of my job, and using those competitors and understanding what they're doing, that's part of it, and being in close contact with research, who have become my best friends. So it's absolutely that. And yes, I also think we're keeping up. Lightricks, Lightricks keeps up with that quality, which is pretty, pretty cool, and I really feel we're at the cutting edge. We have a video model, LTXV, a video model we trained; it's ours and it's open to users, which is also part of our desire to give back to the community and be part of this, of this crazy pixel revolution. And that's it, it's cool: we give a model you can use and get things out of. And I know there was Sora, it's very elite, but in the end I can't use Sora, so I have no idea. And I really have the feeling that we're in a very high place there, that practical place. Yes. Until, basically, are you a technical person, did this come easily to you? Not at all. Originally I started out studying acting at Beit Zvi. I moved to UX because I said, OK, I want a job that actually has work in it and that I'll get by in life, because they made me understand that acting is hard, really hard. And then, and then I moved to product management, but really from a UX background. Wow. And that happened, we'll talk about it, but it happened inside Lightricks, right? It happened inside Lightricks. I started at Lightricks five years ago. I led some project in another app, and my DM is a pretty amazing person who gives a lot, gives a lot of trust to people. Wait, wait, yes, slowly. You were, you were UX. What do you mean you led a project? I was the only designer at the time. Designers were per feature, not per project; this was specifically a new one they put me on. And I also had lots of ideas, I don't know, I was very, I'm very passionate about it in general: psychology, UX, all those things excite me completely, ever since, though I didn't know it, I mean I didn't know it was called product, but today I totally see the connection.

Eyal David: [14:21] So that's it. So like in the project, not some, right, so like in the current project, you were basically in some innovation project, it happened the same way, and then you knew how to step in and start, what, start bringing things down to earth, start bringing vision. What did you actually do there?

Lior Sassy: [14:37] I only designed, but I had a Figma document titled NIF, "next in future," where I simply put designs of things they might ask me for, with headings like, maybe this feature, maybe this, maybe that. A kind of roadmap, actually, now that I think about it. And I showed it to the division manager, and he got excited, he lit up, he thought it was cool, and I think he understood the potential. And then really, I think about two days later he'd already set a meeting with me. He said to me, what do you say about leading this at the product level? Within a month it became official. I'd been at Lightricks four months with the team, and right away, like.

Eyal David: [15:13] So maybe, maybe he'd already come with that idea beforehand. But tell me, how was your move to product? You say, I was an actress at Beit Zvi, and then that wasn't, that wasn't the direction I wanted to go, I mean I wanted to go, but my surroundings told me no. So how did you end up in UX, how did you end up in product from there? Did you want all that goodness?

Lior Sassy: [15:31] UX, already in my studies I really, really loved it and connected to it. And I also studied for my degree at Shenkar, and I also studied film and video. That's it, so it all kind of connected. At Lightricks we make things for creators, so it really felt like my niche. Yes, there were tons of fears. I was a person with zero technical skills, zero, at the level where I don't know how to operate my phone properly. And it was very challenging at first, it was also a kind of trying to fake it. And I remember the division manager would sometimes write me messages like, you said something silly, that's not right, that's not, that's not the correct terminology. I'd be embarrassed, it was terrible. He was a really central figure, if so, he was a real champion, which is so nice, that someone really believes in you. Also internal mobility here, it, it, it happens, and it's

Eyal David: [16:24] it's pretty crazy. So how did it start, if you can elaborate a bit? Is it something born from that Figma page screen, love, where was that great connection born?

Lior Sassy: [16:34] I think there's a chance he took me under his wing in some way, maybe that's what happened. But also, also, look, there were other things, other factors, just organizational elements that worked out so that suddenly someone would manage a new project. I can also thank luck for that. But yes, today product management seems to me like the most amazing thing in the world, and the fact that I came from UX is perfect, and the fact that I came from theater gives me a lot, and like

Eyal David: [17:03] it really is a profession where soft skills are the most important thing. I believe in that so much. So tell me, really, after five-plus years in the profession, what do you think are the most important product skills in a product manager?

Product skills: leadership vs. attention to detail

Lior Sassy: [17:18] I think it's simply a combination of a few, and the balance between them. I can say there's the side of being a leader and an inspiration: knowing how to come up with the cool ideas and strong intuition, and leading people very strongly. And there's the other side, which by the way is almost contradictory sometimes: that whole place of being very detail-oriented and nitpicky, making decisions carefully, doing research, looking at the market and understanding where you stand. And when you find that combination, it's very rare. It can also be learned, but there are people who manage to hold both of these at some level, and in my eyes that's the foundation of product management. And today too, when I see, say, the junior who'll join us, if I see that one side is stronger in them, my goal is absolutely to connect the other side and create some kind of whole.

Eyal David: [18:19] What, what kind of people are you looking for in a role like this? Can you tell us a bit about the criteria? You gave us the skills you'd look for, what else would you look for?

What they look for in juniors and what the interview looks like

Lior Sassy: [18:30] So really we're looking for good communication, we're looking for analytical thinking, we're also looking for a strong desire to learn. I mean, someone came to interview with us not long ago who didn't know at all what big data is, he had no idea, he didn't know. But the very fact that he stops and says, OK, I don't understand this, I'd be happy to understand, come on, explain, that's one of those things where I say, OK, with this person you'll be able to put more knowledge in, and he'll be gorgeous.

Eyal David: [18:58] And can you tell us a bit, first of all, what level of knowledge you expect people to come to an interview with you with, and if you can even give an example of a question you

Lior Sassy: [19:08] UX, let's say, product-UX related, from an interview, that's nice, but I try to understand, interesting that we test a product manager on UX, because that's actually less what we focus on. We focus a lot on product sense, which is the classic idea of how I get to, say, a roadmap for the next quarter

Eyal David: [19:30] from the very bottom to the very top. So you give some KPI that needs to move, and then I need to break down my roadmap. Take me there. Yes, or even, even more abstract than that: pick an app, Google, you have Q1, you're responsible for growth in searches. Yes, that's it, yes, basically we kind of give a KPI. And how much do you expect, or what prior knowledge do you expect a candidate to come to such an interview with? It can really be anything. I think you can be a waiter and have that thinking,

How the product works and the challenges: market fit and moving to web

Lior Sassy: [20:06] that analytical thinking, and have no connection to product management, but you really love AI and you think in that kind of way, and that's cool and that's fine. We really had people come to us from the army, from courses, from psychology, and from data of course, a lot, and I haven't seen any baseline that's required yet. OK, I think it's still worth getting to know your materials or your products a bit, or some background on the company. Right. Cool. So we'll come back to talking about you for a second. You know what, I'll even go a bit back: if you can describe for us, because we can't see the product, maybe just what it looks like at its most basic level, and then... Yes, so you enter the product, and there's the main entry point, which is to start with an idea. It can be an idea that's one sentence, like I said, an American comedy film, and it can be an actual script: I already have a written script, just make me the storyboard, and you upload it. In that screen you can also say what genre it is, what the target audience is, to really give extra details to our LLM model, which we use to prepare the story, to break down the story. And the last point is the storyboard screen. The storyboard screen really shows, divided into scenes, shot by shot, that ad or film or trailer or whatever. And from there you can simply start editing, whether it's changing the text and writing a prompt, or you go to the shot editor and really edit the thing all the way. Cool, so now can you tell us a bit about your challenges? Yes, so there are many. Our biggest challenge right now is the whole question of market fit, and what I mentioned a bit about where we come in at this stage and where we bring value. And also how we stay at the top of the technology. For example, LTXV, our video model, isn't there because we think tomorrow you'll be able to release a feature film on Amazon Prime, but we think it's important for us to keep the knowledge. To keep the knowledge and the capability, and always be at the start, among the first leading this thing. So that's the challenge we constantly face: whether to focus on the present or the future, that's one. The move to web is also super interesting, but I hadn't been on web before. I'm very used to apps, very used to app users, consumers who work like that on the weekend, and to app flows, to app marketing. I mean, how do I even advertise this thing, that was also an issue. What do you mean, how do you advertise it? Basically, if I do the marketing through the phone and in the end I want to bring them to the web, I create a kind of barrier where, as they say, my funnel breaks, and I need to find a way to overcome that. Actually I'll say that one of the things that was easier is that there really was some organic buzz around this, which was actually harder for us in the past with other products.

Eyal David: [23:29] We missed, it works for us in another product we support at my company: we do the sampling through the phone, and then the output, you get, like, hi, think, so you get an email when it's ready, and then you usually open it from the desktop. Cute. I don't know, it works for us. What a product, I really liked it. What other challenges do you have?

Lior Sassy: [23:48] Or you personally, yes? So we also need to get to know the user, and also to know what, I'll call it, good behavior is. I mean, in the apps I was in, Videoleap, Photoleap, which are more, I'll call it, straightforward, it's very clear to me that uploading videos is great, editing them is great, exporting is excellent. And with AI it's a bit harder, because if the person now, say, changes the seed, does that mean they're satisfied and going with the flow, or does it mean they didn't like the image so they need to change its seed? And our data right now is something we're still not sure about, because we really aren't sure; suddenly there's some increase.

Eyal David: [24:28] We keep it aside as something worth paying attention to, and we don't know if it's good or not, in many, many ways. OK, you mentioned you have lots of friends in research. Now I have another question, I don't know: nine months ago, you're saying they basically pulled you from what you were doing. Did you keep juggling what you were doing in parallel, are you still doing it, are you also maintaining it, what does it look like? So no, today I'm no longer there. There was a period when I did it in parallel.

Lior Sassy: [24:54] It was crazy, I was like, because they'd already moved me to the new thing and I still had to work on the old thing, which is like, I'm also the mom and dad, lots of mixed feelings. Did that also happen to you in the previous episode, when you moved to product? It happens almost all the time, because the environment at Lightricks really is as dynamic as it gets.

Eyal David: [25:16] I think that's also something that anyone who wants to move to product should take into account: they'll need to juggle it for a while, and that's part of the entry fee. Absolutely, and by the way, it's also not only

Not getting your heart broken when direction changes, and data

Lior Sassy: [25:27] juggling between two products. I mean, I can work on something very, very hard, and after a few days management will talk to me and tell me, we realized it's not the direction, or worse, a competitor of ours already released it, so there's no point competing on it at all. And the ability not to get your heart broken is extremely important for anyone who wants to be a product manager. That's true, that's true. Did something like that happen to you that you'd like to share? It happened to me tons, but I love it. I really don't, I think it's one of my strengths. Give me switches, changes, I really love it. I feel I can trust people when they know how to say, OK, we were wrong, it's not the right direction. It gives me trust in the company, in Lightricks, as a place I know isn't afraid to take a step back. So I really like it. Some people are more sensitive, and that's OK too. Those are usually people, by the way, who are very into data and very into detail; that's where I'm less strong, I'm working on that step. Still, tell us a bit about that challenge. So really, I can say, for example, I had a product manager at the previous company who would go into dashboards and manage to read them perfectly and tell the story, and I, say, I'm not always there; I need a lot of time next to the dashboard in order to understand what its storyline is. So today, really, like I told you, I build friendships, and we sit, the data analyst and I, and we try to understand what the story is and break it down.

Eyal David: [26:56] And also, and I won't lie, yes, at seven in the evening the numbers sometimes start to swim in front of my eyes, it's not easy for me. What about using GenAI tools to get the insights, the story? So I do that, absolutely, yes, we do that, we do it a lot. I think it's classic, start practicing, I think it's nice. So let's go back for a second, really: nine months ago, six months ago, they gathered you in a room, said yes, what now, we're doing this, applause, off we go. You understand there's a market. OK, now let's talk about the learning you did there. Look, when we talked in the prep call, we talked about the prep call, we had coffee and said, let's understand what value we want to give here. So basically we want to reflect here the learnings you had, the learning you had at the start of the road, as a non-technical person with AI. And basically, right, that's basically the thing. That too, but I think the

Lior Sassy: [27:53] bigger challenge for me was understanding what these people need. These people, I call them, are the ones who are actually looking to do something real, I mean they have a film or an ad in their head. That's for sure, it's like any product, and I'm with you, but there was also here the realization of, here, I have a solution, how do I screw it in for a second. OK, you're talking about that stage. Absolutely, yes, and that happens a lot with AI, really this whole concept of the solution coming first and let's bake something with it. And I still, I arrived at the stage where they'd really created something internal that helped us. They, that small team of three people, started from text-to-image, some audio feature and some preview; that's what they worked with, and they got to this thing. And even today I can tell you that I do, I think I even mentioned it just now, I go shopping, I go shopping with our research team: they show me new technology and I try to understand how it can add more value. It's constantly trying to understand how these things fit in, prioritizing them, and understanding what's hard about them.

Eyal David: [28:56] Yes, so let's see, let's really take that learning. OK, you left the room, and now, I don't know, they moved you to this team, or the forming team starts doing product work, right? What does it look like from here? Because we said, OK, what exists today is quick and dirty, it needs to be thrown in the trash, that's already what you said. Maybe at the most basic level I can also write prompts that return things that shouldn't be there at all. So how did you actually take it from that stage?

Breaking down the need and writing prompts without a technical background

Lior Sassy: [29:26] So it was really breaking it down: why do these people make a storyboard at all, what's the goal. And at this stage you'd obviously already thought about scoping and pricing, or not yet? No. We had some insanely overblown Monday board with lots and lots and lots of ideas, and first of all we started dividing them into pillars of why we're doing this, what they're meant to serve: do they enrich the platform, are they a different feature, are they an improvement to something existing. Just as an example, I can say that our work with the text model was very early: we gave it a few prompts on how to produce this film, and we worked on that model, and from what there was then to what there is today, it's heaven and earth. We optimized the model and trained it. And today it's also our competitive edge, basically; it's only ours now. Someone simply won't find it easy to reproduce. You can't go to ChatGPT and do this, that was a lot.

Eyal David: [30:25] So that was also one of the goals: what will give us that edge, that moat. Now, that's one thing you did, I don't know how much time you invested in it. Where else did the time go, what other things did you do in parallel? First of all, we had to build the platform from scratch.

Lior Sassy: [30:44] There was no platform at that point. It was like a POC, really code that couldn't be used, code written very fast, which is something we also like. You weren't on web, you said, OK. Yes, no, in general at Lightricks, as a working method, POC is one of the things: we understand that you need to feel these things. So it's something done within two months, but you can't work with it, and you redo it. You also redesign it. It was very rough at the time. We thought it was cool, come on, release it, but behind the scenes there was a UX designer saying, no, it's not there yet, and a developer saying, no way, this code isn't mature, and you understand there's more work here. But honestly, we really worked like donkeys. Also, basically the whole sales and marketing element that exists inside a product, inside a platform, was missing, onboarding, which you probably did, not the sophisticated kind, but

Eyal David: [31:36] at the product level, you're saying, I could input something and get back the storyboard like today, just the quality went up, right? Do you want to tell us for a second about the quality improvement you made there, and also what you learned, if you can?

Lior Sassy: [31:47] Yes, I don't know if I can really give details, because it really is something that's part of the power of the thing. But it was me and the PM who was there; at the start we had no data scientists at all, or people more specialized in the LLM field, so we really built those prompts ourselves. Wait, wait, wait, wait, what? Can you expand on that a bit? You're a non-technical person, how did you go about building prompts, what? That's it, so it started with trial and error, and I learned tons about prompts, it's a whole world. And I really, that's one of the things I can say: I'm a non-technical person, and this I know. What are the significant things, the top highlights you learned about writing prompts? Today they're more, like, the most basic. You know, say: don't ask, command. There's the matter of the prompt's order. For example, I can say about GPT in particular that it pays a lot of attention to the beginning and the end and less to the middle, so that's where you should put the less important details. Another tip: if you write "that was cringe," you know it? If you write "that was cringe," it'll give you the same answer in a much more human way. So things like that, which I learned from a relationship with something that feels like a person to me, like ChatGPT is my friend. And basically you'd take demos, as it were, generate lots and lots of options to see what sticks. Tons and tons of stories. And you also do something and it's perfect, and then suddenly someone makes a children's film and it works like crap, because you trained everything on heavier films, dramas and docs, and you need to start from the beginning. So what did you learn around that? How would you make a process like that more efficient? First of all, there's something very good about getting it out to people already. So, say, if we'd had that understanding earlier, maybe I would have shown it to users for internal testing even earlier, to really cover all the use cases. But I can say it's also constant work. LLM is simply constant work. AI in general changes insanely; what we want from it changes according to what we know it wants, and that's constantly, it's the most dynamic thing in the world.

Eyal David: [34:09] So it's work that never ends, you can't fully optimize it. OK, still, we're staying in the optimization area: what KPIs do you measure? Obviously it's a platform where we sell things, so it's clear what the KPI is, signups, but what's actually the quality KPI at the model level? So we have a few. One of the KPIs

KPIs for an AI product: sharing, retention and time on platform

Lior Sassy: [34:30] that I focus on the most, I call it "showing it to your parents," meaning some kind of share, because we have many kinds of share: exporting a pitch deck and sharing it, downloading a single image, and there's collaborating on a project, which we added pretty early along the way. So anything that lets me share the thing now, and obviously one image is less strong than sharing a whole project with someone now, but each of them has its own, say, score. Also, of course, retention, coming back to the platform, creating several projects. One of our interesting KPIs is time on the platform. I think that could also indicate that it takes time. Exactly, that's why it's interesting: we're improving the time of Ruth, but most of the time we'll still want to see people spending more time there. You feel it too; very, very quickly we understood it's a KPI, because we simply got addicted to this platform; you find yourself working all weekend on something. Right. What about qualitatively, do you go in, look at people's projects, ask questions, what does it look like? Absolutely, we have tons, everything. We have a user-testing platform where we watch recordings where people tell us... Yes, what do you use, your own tool? No, user testing is a real tool, yes. Gorgeous. And we also have UX Research, Shirit, who's amazing and sits with us in meetings, and we really, and our business department also organizes meetings for us. I told you, people. Tell everyone. We have Ari Folman, someone I can mention, who, I didn't know who that is, maybe you'll expand. So Ari Folman made Waltz with Bashir. He's a filmmaker, an amazing director whom I really admire; he served with my mom in the army, by the way. And we have many more partners I can't name, but they're all in the filmmaker space, some freelancers and some at bigger companies. We also have lots of ad agencies that create ads for companies, also pretty big players, and it's cool. And that in itself is a gorgeous use case that's also worth a lot of money, right? Yes, it's an amazing use case, and we're building on it: the vision is that this person can create something with LTX, then take it to a film company and present the idea, and basically get accepted with the LTX product faster. The painful moment is how much effort I put into getting that approval, and in the end I don't get it. And we believe that with LTX, with the higher quality and the iterativeness, I mean you can now play with what comes out in a more fun, light and more creative way,

Eyal David: [37:21] you'll have an idea that comes across better, and you'll sell more ideas. Gorgeous. And are most of the customers in this app, as you emphasized, from ad agencies and partners, or is it more, or is it also, and more, people who simply come from ads and...

The users: the eToro ad, Eretz Nehederet and AI enthusiasts

Lior Sassy: [37:37] Excellent question. We really have both, and lots of AI enthusiasts, not even from ads, people who saw the ad we made with... Can you tell us a bit about that, about that ad? Yes, so eToro, a company we did some kind of business with: we created an ad that's entirely AI through the LTX platform, a kind of collaboration. That ad was shown during the Olympics. That same week, our number of users in Europe grew, I think, by 300 percent. I'm making up the number, but it was very, very big growth at the time. And there really is a lot of organic growth, because people who are excited about AI look for it, simply look for this thing. Also Ofir Lobel, who's one of our partners, presented a sketch made with LTX on Eretz Nehederet, I mean really a video where some... They wanted to shoot some part in Greece, but of course they didn't want to fly to Greece just for one shot, and basically they generated that whole background in LTX. It came out pretty cool. No, I even know what you're talking about. Yes? Yes, my daughter loves it. It was hilarious, yes, a hilarious sketch. Right, so lots of enthusiasts, which by the way is also one of the challenges: how do I give them everything they need while not forgetting that in the end there are people who are much less forgiving. When I talk about Ari Folman and his next film, he won't be as forgiving as that guy on TikTok who saw us, got hooked and says, how cool, AI. Or when I talk about some ad agency showing a client, they don't want to tell them, look what I did with AI; they don't want anyone to notice we did it with AI. Right, I know the Fiverr ad they released now, which is like

Eyal David: [39:24] he probably got heavily scolded, AI like that. I think that's it, the world is heading there too. Yes, gorgeous. Or do you have another insight you'd like to share, or shall we move on, because the episode is full of insights and interesting.

Lior Sassy: [39:36] Right, in general I'm really looking for how AI can connect to my world and make my life easier; for me it's a partner. So maybe my tip, which is also maybe already a bit too late because people are really adopting it, but really just

Eyal David: [39:51] adopt it and explore how it can contribute and what it can do. Right. If I go back for a second to the guy who came to you with big data, with that question, OK, that's a question where you can see from it that he knows. On the other hand, I'd come to an interview with you very prepared, specifically from the AI angle. If I'm coming to a company of this kind, definitely reading and learning maybe things you've published, I don't know, maybe you publish on a blog. How can you follow what you're doing? So, so yes, through

Lior Sassy: [40:18] LTX-Video on GitHub you can absolutely go and see everything we release. And we also obviously have our Lightricks website, where you can also see the open positions for anyone interested. And yes, you can easily learn about us online; GPT knows us well too, so.

Eyal David: [40:36] Very nice. OK, it was a really successful episode in my opinion. In my opinion too, I had fun, surprisingly. If I'm not mistaken, and I really hope I'm not, it's also the first episode with a UX woman who moved to product, officially. Nice. Tell me, from that place and from everything you do and have done, where do you imagine Lior in 10 years? An answer, an honest answer: I'd like to be a CEO,

Lior in 10 years, and tips for juniors

Lior Sassy: [41:07] a long shot, of something related to tech, yes, absolutely, and to combine, though, those worlds of content, to steal a bit from LTX, just kidding, no, but I came from theater, so I'd also like to think about how this thing could be, like, streamlining processes in theater, or actually connecting with the physical. Absolutely, something like that, in two worlds that are super different, I think it could be exciting. Gorgeous. Look, you're on the right path, you're probably doing it, because as we saw, you start with something and then it meets you later on; we saw it in your story. I wish you lots of luck. Thanks. And if people want to get in touch with you about positions, about consultations? LinkedIn, sure, happily. Gorgeous, so we'll put a link in the show notes. One second before we leave, do you have any more tips or something around approaching you as juniors, and what's worth coming with? Basically, what I suggest doing, besides reaching out through the website, through LinkedIn, whatever you like, is simply to come with a portfolio that shows relevant experience, and it doesn't have to be product management experience. I mean, it's actually less interesting to say what jobs I've had if they're unrelated; it's more interesting to say, a project I led as a volunteer, or even something I did in the Scouts, in school, or some group project at university. That gives me an understanding of what kind of person is in front of me. So that's what I'd recommend. So better to come with something relevant than with nothing. Yes, than with something irrelevant, yes, absolutely.

Eyal David: [42:35] Gorgeous, you were great. I think, how fun, we'll invite you again to catch up on how you've progressed, in another nine months. Come on, come on, bye for now. Bye, thanks. Hi friends, thanks for listening. If you found this podcast valuable, you can subscribe and follow us for more episodes on Spotify, Apple Podcasts or any other app. And of course, if you couldn't find us on some app, I'd be happy if you wrote to us. We'd love five stars on every platform, and for you to follow us so more listeners can discover us and find the podcast. You can also find previous episodes on any app or on our YouTube channel; there are links in the description. So until next time, come on, be efficient, and bye bye.