AI-generated movie "Dreams of Violets" director on why he used AI instead of actors

本站添加: 2026-05-30

[00:00:00]A new drama is pushing the boundaries of storytelling and creativity. Dreams of Violets is the first movie created entirely by artificial intelligence to debut at the Tribeca Film Festival. It's a fictional drama about five strangers who witnessed something very real. The massacre of Iranian civilians back in January. Here's a preview.

[00:00:30]Joining me now is Ash Kusha who's the director and producer of the film. Thank you for being here. What was produced by artificial intelligence and what was produced by human intelligence?

[00:00:40]>> Well, thanks for having me. um the writing of the script, the ideiation, the blocking and um I would say the entire creative force to make decisions to put footage, the end footage and edited and as well as the sound and music is humanmade. What you see on the picture, the end rendering of the footage is uh made using AI models. So the imagery mostly is uh I would say the AI component and the reason for that is very obvious is because it's very hard to make this kind of movie about such a delicate subject.

[00:01:15]>> Can you go further into that such a short time?

[00:01:18]>> Why not make this with a real crew with real actors?

[00:01:22]>> Yeah. So there are a couple of notes about that. Um, I had to sit and watch and research around 3,000 images and footages of the actual events, the eyewitness accounts. And in order to create a collage of these locations, it would just take months to even prepare the locations uh have a large budget to allocate the uh the the props and the people and the locations. And I think speed here mattered for me. I'm it was my choice uh simply because this is a very personal story. It needs to be told very quickly, right? It's a very poetic impulsive uh piece of work. So I think I thought for a moment that there's no way for me not being a professional film filmmaker as well to go get millions of dollars to make this film. So it was just impossible for me that this is the category of film that is just impossible considering what the subject matters. It only cost you $2,000 to make, right? And and it's my understanding that this wasn't your day job. You have a different day job. How'd you do it?

[00:02:27]>> Yeah. So, I had to basically make blocks of times where I put things to either render to generate um and then just use my time after work and that's why it took two months. I would have uh done this in three weeks if it wasn't for my main job.

[00:02:44]>> Yeah. So, in a statement to CBS News, Tribeca Film Festival co-founder Jane Rosenthal says, quote, "What moved us was not just the technological achievement, but the emotional immediacy and urgency of the story itself." I mean, it does look Ash very different from a lot of the AI that we see. It does look very realistic. But what do you say to critics of AI who say that this is a slippery slope, who say there are still costs like water and energy, knowing you want to continue making films like this?

[00:03:15]They're right. There are a lot of questions around AI and I think one of them is the cost and I think it's been well promoted to use AI a lot more than it should be in my view. And uh I kind of proved it to myself first of all that $2,000 was enough to make the right decisions to generate the videos and the images. Um and I I think the reason for that is that the work done before and after the film are more important and that is to me the methodology, systems, tools and ways of dealing with the the final generation that makes these new uh forms of movies and content. And I think we need to educate people about that because that is the stem and the seed of the future jobs. So I I heard you in a different interview saying like for example the lighting guy instead of physically putting the lights together on a set would be a consultant and working on how to light the AI image we're seeing. Is that what you see as the evolution of this industry?

[00:04:17]>> Think about light sound. Uh think about the music. Think about all these things that need creative choices. Right. At the end of the day I have a background in making stuff in incredibly fast.

[00:04:29]before AI, I was making like 10 albums a month or something. Uh I'm just acquainted with technology, but I think there are a lot of people in cinema, in music, in art, uh that are creative enough to learn these new tools and then come into these new teams and just take on the responsibility and operate these models. And I think these models respond a lot better when somebody who knows the craft, who knows the creative work uh as opposed to just anyone any any consumer that would pick up these tools, right?

[00:05:01]Because they they wouldn't know where to begin.

[00:05:03]>> Did you hire any of them for this project?

[00:05:06]>> Uh not for this project as it's sensitive and I and I'm and I was really learning about this stuff. So it happened over two months. But now I'm in the middle of the second project where I'm actually getting real people. I'm licensing voices. I'm licensing faces because in a fictional uh setting now you can actually uh give people the opportunity to have um a share of the profit of the film >> uh for the work that they do. I think it's it's a good opportunity for reset and I'm not very pro AI in general. I've worked in the AI research for 10 10 years but I'm I'm not necessarily the most like vocal advocate for uh AI the way it is. But I think this is the beginning of finding the better solutions for uh usability of AI and the utility.

[00:05:51]>> Well, Ash Kusha, thank you for your time.