152 | Achieve Videos’ Business Goals Faster Using AI Storytelling with Jonathan Vardi

0:00Hello and welcome to another live episode of the Leveraging AI podcast, the podcast that shares practical ethical ways to leverage AI, to improve efficiency, grow your business, and advance your career. This is Isar, make this your host. And we have a great topic and a great show for you today. We're going to talk about the creative process, but we're going to talk about it in a very specific angle. If you think about it, when we're creating content, any content, and it doesn't matter whether it's AI related or not for generations, the goal of content that we create is to, in most cases, move people to action, whether it's educational content, whether it's marketing content, whether it's any other kind of content, there's a goal that is attached to that content.

0:46Unless you're just an artist and you're doing stuff because you want to do the thing, most of the content that we create is there to drive people to action. Now it has been done roughly in the same kind of process for a very long time, meaning that process from ideation to to making it more concrete and specific, to figuring out what your goals and how to connect the two together, to putting the building blocks in order to create the content, to creating the content itself, was a process that everybody went to school for and learned from for a very long time. It has been implementing across multiple aspects of content creation.

1:26AI is changing almost every aspect of that process in very meaningful and profound ways. And it gives a lot more people a lot more opportunity to do more stuff. Now, I get asked a lot, like, oh, is AI going to kill AI creativity? Because it can now create all these things that wasn't doable before. And I always say it's the other way around. AI is driving to explosion in human expression and creativity because it enables people to do stuff that they couldn't do before. Like people who couldn't generate images before can generate images. People who couldn't write before can write. People who couldn't structure their thoughts in a way that other people can help them to execute what they're trying to execute can now do that.

2:13And so knowing how to go through a creative process with AI is a superpower that is now available to anybody who wants that. And I know that there's a lot of you who want that because knowing how to do this can allow you to reach a broader audience, connect with people in a more meaningful way, and also drive more business. All are things I'm sure that people who listen to this podcast are interested in. And so our guest today, Jonathan Vardy, also known as Jovardy, has been in the video commercial generation for years. He has been producing and creating video-based commercials for a very long time.

2:53So he knows everything that goes into doing this professionally. But recently, or in the last year and a half, he started doing a lot of that stuff with AI. And I stumbled upon one of his videos on LinkedIn, and I was blown away. And I saw the video. Oh my God, this is absolutely amazing. And when I saw the video, I did not know he's been creating commercials for a very long time. I was just really impressed with the level of the AI video that he created. And I reached out and we started talking. And it turns out he's also Israeli. Yeah, so I'll just jump here. Because I'll say that, yeah, I've been making commercials for many years.

3:28That's true. But specifically AI, I went into it to explore other realms of creation. Anyone who's hearing this can use it for their own engagement or stuff. But the reason I came into that is really to explore the creative side. And because commercials you do, and that's, I wouldn't say my day job, but that's like what I do for a living. And AI, at least for me as a director, kind of gave me the opportunity to just explore my creative side without the need of, you know, big budgets and so on. Yeah, just jumping in there. I agree. I think the conversation today is going to be a lot about how you can fuel your creative juices and use AI in order to actually put them to action, whether you're doing this for a business reason or you're doing this just for fun or for just creative purposes.

4:22I think action is many things. Action could be to make someone think about something or to engage you in an idea or an emotion or to tell a story. That's also, these are all, yeah, buying a product as another thing, obviously. No, but I agree with you. I even said education, right? So like whatever we create stuff, even if it's just, unless you're going to keep it in your basement, but if you're a creator, usually you want to share what you're doing and it could be to provoke emotion in somebody. That's still fine. It's still creating something with somebody else's. You're impacting somebody else one way or another.

4:59Yeah, definitely. Definitely. But the reason I am saying that is because if you think about the commercial process and working with clients in a traditional way, and you try to implement that kind of process with AI, that's not going to work. That's going to be a total suffer from all parties you need when you go into making those stuff obviously you need to have a clear vision in your mind of what you want to do but also know the limitation know what you cannot do be open to see and you cannot obviously you can come and explore and i see many people who are doing great stuff and i know that they are more experimenting and checking which tool is doing what and what prompt drive what when i come to do something i'll usually have a very specific script, very specific story in my mind.

5:50Not that I don't go away from that in the process. I do. Sometimes I see a great angle, which I didn't start off, and I'll say, okay, that's amazing. I want to keep that. And maybe I'll explore more on that. But I need to have a very clear vision. And I think this is true to anyone who's going to do this tour, a very clear vision to go in. and even clearer vision than you might need when you go on traditional shoots because there is no limit. And you can literally lose yourself in prompting and waste so much time. And to be honest, the shots that I spend too much time on, this is where I stop and say, wait, do you really know what you want to, do you know what shot you want?

6:35I'm asking myself. And usually I'll say no. I don't want because even if you don't have a clear picture of the exact thing, you need to know to spot when that's what you wanted. And in AI, that's more important than in others because of the open, because of the options you have. I love what you're saying and I want to touch on it because I do this a lot, right? So I love creating images with AI tools and I do this for everything that I do, right? I don't have any stock photography licenses anymore in any platform. I just create images for presentations, for workshops, for literally proposals, everything, and sometimes just for fun.

7:17And I do find myself just like you, where sometimes I would spend 20 minutes on one thing and then you go back, isn't one of those good enough already for what you need? And I'm like, yeah, it is. It's just not, it didn't trigger the emotional side of me to say, oh my God, this is exactly what I wanted. But in many cases, you're like, I actually, the 12th one out of the 57 that I created was actually close enough. And nobody else will know, right? Other than me, that it wasn't exactly what I wanted. But that's what you're saying now is true, not only to AI, because like when I'm on set and sometimes we have 200 people and every minute costs, I don't know,$10 ,000, the most important thing to say, we have it, we got it.

8:02And you never have it. you never got it like you can always i can keep on shooting every shot again and again and again i just need and i'll never get my shot i just need to have it good enough to move on and again that's when i'm working and doing those hot films you need to commit very fast because definitely like if you're making just images that's one thing and then your final product might be the image or like a series of images and that's one thing but if you make a video then those images are then fed back to the to different engine and that's going to drive the video so once you say okay that's that's that's image i'm going for you gotta lock that you can't go back and change that because changing that uh and you'll see maybe later i'll share a bit more because right now i'm doing you do the image and then you go to the video prompt and now we saw which has its enhancement it goes it on another layer.

8:58So going back is very hard. Yeah, because it breaks the rest of the process. So really, you know what, this is a great segue. I will say one more thing from a personal creative perspective. And then I really want to hear, start with a high level of your process and then start diving into the details. But still some of the favorite, my top favorite AI images that I created, and I created thousands are still the ones that I actually invested a lot of time in because I knew exactly what I was looking for. And in some of them, I invested, I'm not going to say, but too much time getting to exactly what I wanted for, but I got it.

9:39And now they are literally the pinnacle of my AI creation. And they're my favorite slides in my presentations and so on, because I nailed exactly what I wanted. And that's another thing that I want people to think about because people are saying, oh my God, you invested an hour and a half in creating this one image. I'm like, okay, that one image couldn't have been possible before. And if it would have been possible, it would have cost me$20 ,000 to create because it would have needed to shoot it with an actual camera, with actual actors, with actual lighting. 55 times people cost you a lot of money and then edit that and then Photoshop it like it would have cost me so much more time and so much more money.

10:20So yes, I invested an hour in it and now it's an amazing output that I can use across multiple things that I'm doing. So I think that also needs to be taken into perspective. And that is tenfold more correct when you start dealing with video, like doing stuff with video with AI, even as I needed to render this 25 times before I got it right. I'm like, okay, compare that to you having to actually shoot this in an actual shoot. That's still free. Yeah, yeah, yeah, yeah, it is, it is. So with that, let's jump into your process. Let's start high level. Like when you're going to produce something, what is your process that you're taking both from a thought perspective as well as from a tools and practical perspective?

11:01sure uh i'll already share just so we have some visual so and it really depends and it really evolved okay because i think the first ai proper ai video i've done some experiment but the first proper ai video is this one which i'm not going to play it but you can see some gifs out of it And this one was done about three months ago, three or four months ago. Obviously, I was playing with Gen AI tools way before that, also with Gen videos and all that. But this is the first time I said, okay, I'm going to dive in and I'm going to do something complete. And that was done in the Gen 48 competition.

11:45They had the two days, 48 hours where you need to create a show. And what I like about this is I started experimenting with it and I felt that the image to video doesn't really give me enough life, the kind of life that I wanted to. It felt, at least back then, more a bad filter or like a good filter rather than like really something that is alive. and the text to video not the image video the text to video did gave me that and all of this video was done yeah and i think that it is very interesting and still today when i'll need something like a simulation or something that is really moving i don't know a snowstorm or leaves flying in the air or stuff like that i'll always go i'll never start with an image i'll go with text to video and i'll try to find a text that that suits that because the image to video is it has these benefit and you can get consistency of character you can get consistency of style but you cannot get the same life that you see here that's that this is really i'm amazed by the way that those obviously we all are made by the way it's literally a magic um and the way i did this is i created the i don't have the prompt here it's two two thousand light years from yeah Yeah.

13:14Yeah. But I can tell you the structure. So I start with the base prompt that basically talk about the world, the color palette, what type of camera movement and what type of camera I'm using. And this kind of defines the world that I'm in. For example, here, I'm in the forest. The colors are like green. There is some green tilts in it. The camera, you can't really see it on those GIFs, but a lot of the camera, a lot of the shots have this kind of a light, like a flashlight. By the way, for those of you who are just listening to the podcast, what we're looking at right now is as if somebody's walking through the forest at night with a flashlight, and it looks completely realistic and lively.

14:01And it's really incredible. And when I saw this, by the way, I wasn't sure if that's AI or not. It's at that level. because, and I think part of the magic of it is that it's a night shot with a flashlight. So by definition, it doesn't look like quote unquote real life because we're not used to seeing it regularly. And I found that video, I saw it way before and it's absolutely magical. And what I wanted to explore here is really texture because I wanted to see how texture feels and how leaf and the magical glittery stuff or glittery animal. So it was really like an exploration with prompt, but an exploration of materials.

14:43And again, so yeah. Quick question about this. So again, guys, what we're seeing is a pretty long video. How long is the whole thing? I think that's about two and a half minutes. Two and a half minutes. And all in it is like a night shot in a forest with like different glittering creatures. so it's as if creatures have these diamonds that are covering them but some of them are black some of them are white and they're shiny and they pop up when the light hits them and it's really interesting how many attempts did you do of short little segments to create this two and a half minute long thing and then what was your process i don't know how many the process was first element of the process was okay i'm committing for this meaning i'm going to start and i'm going to finish the film i didn't had again we'll talk later about different kind of project but in this one it was just like okay i'm going to make it doesn't matter i didn't have here specific story i said i'm going to explore and i'm a documentarist in this i'm diving with my documentary perspective mind into the ai world and i'm seeing what's coming up and straight away i wasn't collecting the images on my folder.

16:04Any video that I liked, I immediately put on the timeline and started connecting them together. Okay. And this is amazing for, by the way, any project I do with AI, I do that because of many reasons. Some I'll touch a bit later, but I write the story, even if I have it. And sometimes I work with a storyboard already. Even if I have all that, I'll still rewrite the story on the timeline with the shot that I'm getting. So how many prompts? How many attempts? Let me pause you for a second. Let's talk tools. Like which video generation tools did you use? And then where, when you say you dropped them on the timeline, I assume you use like a professional editing software, but I still interested you.

16:45What's the process? Yeah. Yeah. So this one was done completely with runway. So all the prompts, all the video generating was done with runway. on the Gems 3 model and then music with Udeo, which is another, like it's an AI music maker. And then voices, again, I'm using 11 labs. Some of the voices are recorded. Some I use the 11 lab for the, again, another AI tool where you can create like sound design. You can create a character. Yeah. Cool. Okay. So the process is, in this particular case, you come up, you test different things, you grab the videos that are generated, you drop them on a timeline in an editing software like an Adobe Premiere or...

17:34Correct. In this specific point, again, this is a very experimental piece. The story is built while you walk, while you go. And the more you spend time on it again, the more you craft the world and more stuff and that way the the material tell the story i'll share later different kind of stories they are more yeah more story driven more character based and there the process is totally different and i can for example if i'll go the film i was made straight after that so after i finished genesis i took a few weeks off i did some commercial projects and then i literally locked myself i don't think you can see i don't have gifs of that so i'll just play it with on mute yeah i quite like the project well again i'm very proud of it and yeah here you can just see it in the background this is actually a story i wrote i think two two and a half years ago and i actually went to some producers and some company that reps me with it trying to maybe gather a budget to to do it and everybody just threw me off the i don't know how to say threw me off the stairs the steps they just basically told me john nobody would give you that that's a three million dollar project and nobody in the world would give you that amount of money for a five minute film write something simpler please and we'll help you and I didn't draw anything simpler but a year and a half later or two years later I felt that the tools are already in the place where I can again maybe not make exactly what I want but tell the story so obviously I had a lot of yeah limitations and and stuff that I had to do to tell in a different way with the limits of the tools but I could tell the story.

19:35And here you have consistent characters, you have consistent environment. And that's obviously 90 % of this project is image to video. And in here, and most of the other projects that I will share later, the process is very similar to making a film in a way because before I wouldn't jump and make a shot before I'll make I'll go and do any shot I'll gather all the sources meaning I'll create my character design I'll create my the room design I'll create if there is a specific element in the shot I'll create that and then I'll gather them together and I'll use for example the character I use me journey mostly except if there is too much blood in the shot and then i'll use flux because me journey will block you but or you can try and you'll see soon there is a lot of blood here you can just write i don't know dark red liquid or stuff like that to say yeah to trick the model yeah it's not blood coming out of your brain that's just like red very dark red juice so what was i saying yeah um okay so let's let's go a step back.

20:55So in this particular case, you had a very clear story in your head that you wanted to tell and you went and created each component on me journey. And you said separately the characters, separately the background, separately specific elements that you wanted. Correct. Yeah. So, but again, I'll just stop for a second because it is also a lot about the way you direct it and how you trick how you trick the audience to feel as if they are in the same world for example here i built this world where you have a lot of like fences and this harsh light and so it doesn't and this concrete so for example this shot i just have this foreground and background but i feel we're in the same place so i don't need the entire world i just need the elements that connect me to the to where i was and again that's exactly like when you're on set you'll you'll light first you'll do the long shot the establishing shot and you're going to light that and you're going to set it up doesn't matter maybe you'll start shooting the other because the planning of the day is you have some other things that you need to do before but you'll start setting up the long shot because it's much easier to fake the close up later you can do it in a different place even.

22:15Yeah. So two things about that. First of all, for those of you who are not professional in this, and this is a little confusing, I think the important thing here is to have a vision of the output that you want to get, either because you saw an example or because you just have it in your head, and then create a consistent universe, a consistent environment, a consistent style across the different shots that you're having. So in this particular video, and again, those of you who are not listening, it's a very dark, both from the visuals, like it's literally dark, but also in means of the theme of it.

22:54It's like a dark theme, like environment, but all of the shots have like the same style. It's like a dark room. They're very dramatic. There's very specific lighting point in each and every one of the shots beyond the dark background like it's all very consistent and you can create that image journey do you want to dive into this a little bit like on the actual how to create a consistent images on me journey that then can be a starting point for a video cool i can do that and let's see if i have would i have any journey here

23:32because i think that's part of the hardest part of the trick is getting consistent images in mid-journey that then can be the starting point for videos on runway or cling or now Sora and Vio. These tools keep on coming out and upgrading all the time, but you have to start with like solid, consistent images. Otherwise it just doesn't work. Yeah. Okay. So I'll share, I think this is out of Blue Star, the video you posted or you reposted when sharing this again i did a lot of spelling mistakes and so you'll probably see that as well in the prompt but you can get the wall let's see so this is like the side again you have a lot of images to create anything and then i'll take one image that i created and then fit it back in and create that so it's a very iterative process i'll never get the results i want on the first round but at the end I will.

24:34I want to say something about what you just said because I think it's important to anybody who's creating anything with AI whether you're creating a blog post or an email or anything. Most people stop too early meaning they will try once, twice, five times it's not what I want, this is a waste of time and they're going to go away. And what you find is that first of all you develop a muscle Like something that will take you 50 tries the first time will take you 20 the second time will take you 10 the 20th time. But regardless, once you learn how to do this, you can become better and better at reusing stuff, reusing prompts, reusing processes, reusing images to get to the outcome faster and more accurately.

25:21and what we see here and again those of you who are just listening and not watching we looked at me journey there were hundreds of images many of them are the same but not exactly the same in order to create that universe that this is like a star wars kind of theme yeah and again i'll say like i suppose what you're saying say that like when you make a film the image is just like one part of the process and yeah any image most of the images i'll do before i'll even put them on the clean me journey for clean runway so whatever i'll i'll take them photoshop i'll clean all the stuff that that needs to be clean and 99 there are a lot of stuff to be clean because and you want to clean them on the image level because otherwise you'll need to clean them on a video level which is a lot more work so they'll go to cleaning and then they'll go to upscaling uh with like topaz 8 giga thing or any other magnific or any other uh upscaling that give you a little bit more detail and then you go to the videos and so on yeah don't expect writing a prompt and getting a video that's not gonna happen but if you're committed to making it i think again these are crazy tools and at least the industry that i come from they are doing a lot of changes yeah so let's read i'm going to read the prompt with your permission just to give people an understanding what goes into a good prompt to get something and i don't know if you started with an image on this one or you started fresh so in mid-journey those of you who don't know you can make an image as part of a source to the input to the yes oh it does there There is an image here.

27:06Yeah. I think here, no, you can read the point, but this is actually an interesting reference because what I did, I had the long shot of the spaceship and I just did a screen cap of the spaceship that I created for the long shot and just fed it in. It's not that I wanted this specific thing, but I wanted it to take some of the color, some of the elements, and then that also helps with consistency. Yeah. Yeah. So again, just to describe to the people who are listening, there is a very zoomed in version of a part of a spaceship and kind of like the flow of, I don't know, plasma on top of it, and it has that bluish glowy style to it in a dark gray spaceship.

27:47And then you use that as a source when you prompt the other one and you can define it to use that image as a reference and then write the next prompt. So if we now go to the next prompt, the prompt is saying a cinematic 10 millimeter lens close up shot capturing a sleek metallic wing of a futuristic spaceship taking up a significant part of the frame. The wings, sharp edges and intricate details are illuminated by the glowing blue light trails of faster than light travel. So there is, and this is half the problem. I'm reading you this so you understand what goes into this. People say, oh, give me an image of a spaceship flying fast.

28:32And they're not getting the level of details that you can see in this image. And what I tell people is it all depends on how much you put into it. These tools are very good in understanding what you tell them if you tell it to them in a clear fashion. And you have to think about, if you had to describe this to a different person and have them be able to sketch it, how much detail would you put into it to have the other person sketch it properly? Do the same thing here with the benefit that you can give it a reference image to start with. And so this is amazing. Obviously people are much easier.

29:08And also I'm really trying not to add any like IP references, not in the prompts and not in the images. First, because I just feel it's more fair. I don't want to write a spaceship style Star Wars. I'd rather write because then I'll get a Star Wars kind of spaceship. I'd rather write the colors and write the elements and write like all that. If I'll have just like an art director with me, I might actually tell him, listen, it isn't Star Wars spaceship. Can you design it? And he'll get what I'm saying. So, yeah. Okay, so step one is having a story in your head. Step two is creating those images in mid-journey or Flux or any other tool that creates high-quality images.

29:51Step three, you said, is removing any artifacts or issues. Yeah, just like basic cleanups. I don't know if here I can show anything, but every image has a lot of artifacts. Stuff that not necessarily needs to be there, especially if it's AI created. yeah and then you go that you go that yeah you use that as a source now i'm using all the engines at the same time so i'll use i'll always throw one image to everywhere and and i'm using the i have a subscription in like each one of them individual and i'm also using a wavy which is a great tool that just gives you a node-based place where you can create your own workflow and here for example again from the same project you'll see again this is a prompt for the video so you'll have a cinematic shot speed of light gopro shot flying hyper fast blah blah blah living trails of blue light behind it energetic 10 millimeter what film stock and so on and then I'll feed it to cling which is one model and see what I'm getting this is I think just text to video.

31:04There's no image here. And you can see that the style is the same because here is after many prompts and many attempts. So I get the same feeling. And this is minimax text to video. Same, but maybe I use some of it. I don't think so. And then here I have minimax. So I have the flux images, which I've created. And I think this is quite interesting because you get these images. I don't remember which one I've used because obviously there are, like you see many, sorry, many attempts here and stuff. So we'll just zoom in. And again, this is just for one shot. So what's interesting here is I start with this image.

31:48Sorry for the people who are watching this on. Yeah, so again, what we're looking at is this canvas. And in the canvas, there's all these inputs and outputs. So the inputs could be like a prompt or an image. and then it connects with lines on the canvas to the different AI video generation tools that then gets fed both the image and the prompt, but you can feed it to multiple tools at the same time and see if you're getting different, better, more appealing outputs to what you need. And then you can pick and choose the mix and match once you do the editing in the end. The tool is called Weavy.ai.

32:23So it's W-E-A-V-Y.ai that allows you to do that. And what we're seeing right now is an image of kind of a tunnel of light and then the video that's generated from that yeah but i think what's interesting here just a second is that see what's happening to the video these are the shots i wanted for like this very fast light trail kind of POV from the spaceship. And I've created those images. Okay. Now, I imagine when I wrote the prompt and when I made the image, that the camera would go forward and dive into this image. Yes. But actually, it didn't. This is a different image. It goes somewhere else.

33:12Something else happened. Okay. Yeah. I don't know why. And here, there's a moment that I took. and it did took the elements and like the color and the vibe and everything but it made something that is like half the image and half the text prompt which again I think is like similar to what I read before and this is the same so yeah you go no so I want to say I saw this so yours is even more expected because the video doesn't know what this is. But I had it happen to me when I had an image of a person that I created in Midjourner or something, a person walking, and he was walking down like a street and I wanted to see him walking.

34:04And when I fed it to the video engine, it said the person walking the street, blah, blah, blah. And he had the person walking backwards. And I was so surprised. I'm like, how does that make any sense? But it doesn't know. Like you told the person walking, you didn't tell if it's forward or backwards. You assume he's going to walk forward, but it's not necessarily the case. So what I've learned is that when you want to have a clear movement in those video tools, you've got to describe very clearly the direction of the movement. And even then, it's not guaranteed, but it increases the chances that it's actually going to do what you thought it's going to do versus what you assumed it's going to do.