Loading summary
Transcript1 lines
- [00:00]
Host
On my podcast, I talk a lot about what is going to happen with AI in video, some of the advancements, some of the tools, and overall, I'm very, very bullish and excited on this. People being able to direct their own movies. There's a lot of really exciting things here. Now, that being said, one perspective I do not often share is the perspective of filmmakers. What are people that have dedicated 10, 20, 40 years into this industry and have blood, sweat and tears working their way up? And film studios, what do they think about this? Because in a large way, film studios, set designers, a lot of this goes away with AI and as AI gets better, and I know there's always going to be those roles and there's always going to be movies probably made with humans, but there's going to be some massive shifts and changes. And if all the top movies are being generated by AI, or people are starting to make more movies and the watch time is starting to get more democratized. So it's not just big studios, big production companies, Warner Brothers, Disney that are making all the hits, but like a lot of hits are being made by regular people. I think this really does shift the way things are. Things are going to change. So what are the, what are the movie experts saying about this? What, what's their opinion today on the podcast, I want to break that down and talk about a really interesting panel that just happened at a movie convention from some of some incredible movie creators. What their thoughts are, and they had some ideas that I had not thought about before. There's some ideas I would like to argue with. Let's break all of it down before we do that. If you haven't already, it would mean the world to me if you could leave me a review on this podcast. It really helps me get found in the algorithm. So if you're on Spotify or Apple and you haven't done that, I really, really appreciate. You know, on Spotify, hit the about tab and drop some stars. Now, if you're on YouTube, if you can hit the like and follow and subscribe, that would really help me out in the algorithm as well. So let's get into the podcast. The first thing I want to talk about is where this whole conversation recently took place. So there was a number of people on a panel at a conference called Siggraph. This happens in Denver. There was Nicola, there was Freddie Chavez Olmos from Voxel Studio, Nicolas from Wonder Dynamics, and then they had Michael Black, who is from Mesh Kapad Max Planck Institute. Okay, so all of them essentially are talking about the potential of, you know, AI and how it's going to change systems and how it's going to affect media and et cetera. So what was interesting to me was their consensus when they were talking is that of course these tools are useful. They're like, we can't argue they're not useful but and also they, that, you know, they're changing very rapidly. They're like, these things are obviously useful. They're changing very rapidly. The innovation is very fast. We need to prepare for radical changes is kind of what they're saying. But they're saying that these tools are not the immediate future. Now, I don't know how immediate you want to say, you want to be like the end of the year, the end of next year. Like I think these things are, this is happening very fast. One of the things that they kind of talked about first was. And so you had also have to kind of put yourself in their shoes, like they're trying to justify their existence. And I'm not saying that in a bad way, like they're justifying it like their existence isn't justified. I'm just saying that's what they're doing. Right. They're trying to explain why they're better and anything. So it's just interesting to me. Right. And I'm sure I do this in, you know, if you're saying how AI is going to replace software CEOs, I'm sure I'd be like, no, but wait, wait, I'm. Or podcasters, I'd be like, wait, no, I can do things that can't. Right. So they're doing that. So. But it's interesting to see where they come from because I agree with them on some points, but not on all they said. One of the first things they kind of talked about was what they call the impractical nature of today's video generators. Okay, first things first, I want to say today's video generators aren't 100% incredible. There's a few that are pretty decent. We're really waiting on OpenAI Sora come and wow us all. And there's a handful of others that are out right now that are doing some, some impressive things. But torture Vic specifically said, quote, the misconceptions of AI that it's a one click solution, that it's going to get you a final VFX shot and that's really impossible. Maybe we'll get there, but if you don't have editability, that black box doesn't give you much. What we're seeing Right now is UX is still being discovered. These research companies are starting to learn the ways of 3D and filmmaking terms. So essentially he's kind of pointing out that the language you use to generate something is a really aspect of visual creation. And this is something that I personally have said a lot. If you are a professional video creator, you know, it's like, okay, I want you to pan from right to left and have like a slight zoom and fade in the shot, and fade out the shot and make a start on a black object and move to a black object. So for the next scene, we can splice it this way. Like you have all of these things that you're going to be saying that an amateur isn't going to, A, probably be thinking of half the time, or B, be sane. Now, to say that amateurs won't be able to essentially create really powerful stuff, I think is you're underestimating, number one, the ability, if you have a generally available tool like Sora, whenever it launches, that is able to do stuff. I think you shouldn't underestimate that, number one, people can learn a lot of the filmmaking terms because this is what happened with midjourney. Midjourney was very, very much dependent on having like, a lot of photography terms in there. Like, I, when I'm using Mid journey or originally, and they've kind of fixed some of these things, but I would say, like, shutter speed and ISO and frame rates and like all sorts of like, crazy things. Not frame rates, but all sorts of, like, crazy things that I would just find other people be like, hey, I created this image and this is the prompt I use. And they have these amazing prompts and you can learn from their prompts how to replicate it. And then after that, every time I want to create something that looks like it's an actual photograph, I would say ultra realistic image when I'm, when I was using midjourney. So there's like all these little tips and tricks. And so I think just being like, oh, like everyone doesn't know how to use all the video stuff, I think is a little bit naive to say that, like, if there's a widely available tool and using video terms is going to make it better. It's going to get shared within a matter of six months. Most users, or a lot of users are going to know how to essentially prompt it in specific ways to get to impressive stuff. So I would say, like, don't. And then also the software does get better, right? So. And I compare all of this to video image by the way. Because we're seeing kind of the same thing happening with like, the evolution of like, Mid Journey to image generators today, where that was kind of really important at the beginning of Mid Journey. Mid Journey has made a lot of adaptations to their prompts. That's not as important. Dall e completely got around it where they write their own prompts for your dolly. And it's kind of a little bit nerfed, but it does that. And then you have, you know, Elon Musk's GROK has just added Flux, the open source image generator, to their model. And Flux, you just, you tell it, you know, you just say what it does and it generates like an ultra realistic image of like, anything. And it needs way less of these. So there's ways to definitely design these to need less technical terms. But in any case, it is a good point he brings up, but I think he's underestimating how fast people will be able to learn these. So here is what he says about this specifically. Quote. I mean, things like yoga poses, ballet poses are some classical things we have names for that we can define. But most of the stuff we do, we don't have names for. And there's a good reason for that. It's because humans actually have inside them a generative model of behavior. But I don't have a generative model of images in my head. If I want to explain what I'm seeing, I can't project it out of my eyeballs. And I'm not good enough of an artist to draw it for you. So I have to use words. And we have many words to describe the visual world. But if I want to describe to you a particular motion, I don't have. I don't have to describe it in words. I just do it for you. And then your motor system sees me and is active in understanding that. And so we, I think it's a biological reason, a neuroscientific reason that we don't have words for all of our emotions. Okay, interesting point again. Right? So he's pretty much saying, you know, you're in a movie and you're. The actor's not doing what you want. You're like, hey, man. Just like when this, you know, when the bullet shoots across the scene, make sure to swing your right arm up like this and duck your head and act shocked and like, you know, tremble a little bit and like move your arms in like this kind of motion. And you could just show them like this kind of motion. And they get it right. They see it, they're like, okay, I'll do that. He's saying, when you're making a movie, you can't exactly tell it that. Now, my argument for this is if this really is a fundamental part of the moviemaking process that you can't get around, we're going to do what we did with images, which there got to a point with images. And actually we'll go even before that with text. When I want ChatGPT to write in a certain tone and style, I have the hardest time in the world explaining what that tone and style is. And sometimes you get so frustrated, I literally just go copy, like an article I've written in the past or a blog post or a caption, and I'm just like, copy exactly the tone of style I used here, or a different person copy the tone of style they use now. If then it's able to do it. So we, we upload an exact copy. We do the same thing with images. Now there's a lot of image models where we upload an image and you're like this person make them smiling or like change, alter this image slightly, but we give it the image. And I think if it really was that big of a deal, we do the same thing with video. So it's like I want, you know, this character that I have playing this role in my film. I want them, when the bullet comes, to swing their arms like this and you like, record a video of you swinging your arms like that and you upload it and it will be able to follow you. So I think that there's going to be some interesting things like all this to say. There's obviously been software that has helped with like animation doing this kind of stuff for a long time, but we're moving to a point where it's like video, image, text is our inner, is our input and output. And I think that's kind of interesting. So you see things like the visual. The visual AI elements from ChatGPT Vision, right? Where it's watching a video of you. And I think a lot of that is actually just screenshotting and taking in images. But regardless, you could take five. You could take five images, which is like a slow frame of you making a movement with your arms, and then guess how the images link together with an AI model, right? Like clip them together to turn into a video and know how to make the character in the video it's generating move. So I think there's a lot of ways to do this. And I think that if that really is like a critical point that, that will. That's the solution right there. So I don't think the whole we can't show the model what to do is actually an issue. So all of this is quite philosophical and I totally get it. Nicola, Freddie, Michael, all of these are amazing, you know, artists in their own right, movie production people. Now there's one interesting point that they bring up is kind of interesting and I have a bunch of takes on it. So Chevron is almost. He is from Mexico and he essentially was kind of talking about how he had to move from Mexico to la. All of the obviously money and expertise is concentrated in LA when it comes to video production. He said, I had to leave Mexico because I had no opportunity there. I can see now having that same opportunity for people who don't need to go overseas to do it. So that's kind of, you know, he's saying this is opportunity black. So that is. Michael Black, though, is worried that sudden access to these processes, he's, you know, might be. Have consequences we don't understand in the short term. So he said, you can give somebody powerful car that doesn't make them a Formula One driver. Right. That's a little bit like what we have now. People are talking about everyone's going to be making films, they're going to be garbage. Quite honestly, the democratization thing is exactly what Chavrez said. And the power is that maybe some new voices will have an opportunity that wouldn't otherwise. But the number of people making really good films is still going to be small, in my opinion. Okay, I get what he's saying, but I still think they're underestimating people. Like, you know, the best photographers are like, look, if you give everyone an image generator, most of them are just going to be creating garbage images. Like, I've seen so, so many incredible images coming out of these things. And people try and they hone their skills and they get better and they work at it and they perfect something and they make something amazing. So I think that, just discounting everyone's images are going to be like, crap is probably not super accurate. Like, yes, when you, your first one you ask it to generate might be garbage. You're going to hone it. You're going to try to make something amazing. I really believe we're going to have millions more people making incredible films with these tools. So I don't think it's going to be just a handful of people that are still amazing. All right, so, yeah, anyways, they have a bunch of other. They have a bunch of other things. The last point I wanted to make, the last thing that I thought was interesting was from Chavez, who's talking about AI generated images and how they have kind of some of the, some of the innovation we've seen in like. It was specifically kind of referring to like some animation. So he says it's, it's going to have the same, I think reaction that people had when we saw the first Final Fantasy Movie or the Polar Express. Something's going to not be quite there yet, but people are going to start accepting these films and instead of a full CG film, it's going to be a full AI film. And I think people are going to see, even at the end of this year, people are going to get past that, like, okay, this is AI and people are going to accept that. Okay. I personally think that this is very true. The Polar Express, if you go back and watch that movie, is a movie I watched every Christmas growing up. I don't know why I loved it, still do actually, and make my kids watch it with me. But definitely if you go back and watch it now, I think it's like 20 years later. Yeah, there is a whole bunch of like, oh my gosh, this is really old animation. At the time it was incredible. I think that there's going to be something people get over. Something that Black said that I think is a really good point is he said Pixar taught us this very clearly. It's all about story, it's all about connecting to the characters. It's about heart. And if the movie has heart, it doesn't matter if the characters are AI I. I think people will enjoy the movie and that doesn't mean that they're not going to want human actors. There's an excitement in knowing it's real humans like us. But to see a human at their peak of their game, it inspires all of us and I don't think that's going to go away. I agree with him there, but I 100% also think that there is going to be so many movies that are just really, really incredibly made that are going to be, have amazing storylines. You're going to connect with the characters and that's what's going to be really important here. Um, this is what people are going to sign up for and if it's AI generated and maybe it's not a hundred percent like a studio film. My wife loves watching plenty of B rated romance comedy movies and I watch them all with her to make her happy. And they don't have professional actors and a lot of the lines are cheesy, but, like, she loves them and it's whatever. I think it's going to be the same thing with a lot of these AI generated videos. And so I'm really excited to see where this goes. I think it's going to be huge innovation. And I would say definitely don't, don't, don't assume or don't count out the innovation and the excitement of just average, everyday people creating amazing movies.