Hollywood Green-lights Its First AI-Only Films | Gorkem Yurtseven

Gorkem Yurtseven (0:00)

Can a single person productively spend a lot of tokens doing their work? Only coding agents have this in the AI market. No other AI sub-industry can spend this many tokens as video generation or coding agents. And this is happening because people are now sitting in front of a computer at work and generating video all day long.

Sophia (0:22)

Hello, everyone. I'm Sophia. And today I'm here with Gorkem Yurtseven, who is the CTO and co-founder of FAL. And we're actually here live in San Francisco, as you can see by the beautiful, we have the beautiful bridge in the background right there, because we're here at the Generative Media Conference here in San Francisco. So I would love to get your perspective as actual co-founder of FAL. What brings us to the Generative Media Conference today? What is the whole point of this conference?

Gorkem Yurtseven (0:49)

We just kicked off the keynote and I was able to watch the first session. There's an incredible amount of energy outside everyone.

The community mostly lives online for Generative Media, and this is an opportunity for everyone to come together, meet in person for the most time, for the first time. So everyone is just very excited to see each other. We have a really good mix of individual creators, people who are passionate startup founders, but also executives from studios. We just heard Sean Bailey, who was the president of Walt Disney Studios, now he has his own studio. So it's a really good mix of people coming together and talking about what's next in entertainment and filmmaking and advertising. It's a very diverse set of people.

Sophia (1:38)

No, it's fun to see, because I don't think we typically see all this diversity when it comes to maybe an AI model conference here in San Francisco. We have actors out here, we have Disney execs out here, so it's pretty cool. I want to get your perspective right now when it comes to Generative Media.

Where was it maybe a few years ago when FAL actually started? Where is Generative Media today?

Gorkem Yurtseven (2:01)

So I would say, let's not go back that far. Maybe last year we had this conference around the same time.

We were talking about the future, we were talking about some of the very short clip use cases. But since then, it really reached professional level. And we call it like token market fit. In our definition, Generative Media has reached token market fit. And when I say token market fit, I mean, can a single person productively spend a lot of tokens doing their work? Only coding agents have this in the AI market. No other AI sub-industry can spend this many tokens as video generation or coding agents. And this is happening because people are now sitting in front of a computer at work and generating video all day long. And this is a productive use of their time because it's still much cheaper.

And then it creates a new medium, things they can do, things they couldn't do before. So they are willing to spend all these tokens doing their work. And this is the biggest change in Generative Media in the last year.

Sophia (3:09)

What would you say is the hardest part, though, of Generative Media? Because, of course, we're seeing this when it applies to coding agents because it's verifiable. You can say, does this code work? Is this correct? Generative Media, that's a whole slew of things.

Gorkem Yurtseven (3:23)

The hardest thing, and it has been from the beginning, is controllability. Like people don't want to treat these models as slot machines. They want to think, they want input into the model where they can control the output a little bit better. But that complicates how we interact with these models a lot. Because now, not only you have to write a very long prompt, like the prompt itself becomes almost like an API. It has its own syntax and things like that. But also, the inputs are getting more complicated. It's like now very common to have a low resolution blender input that now becomes part of the prompt. For that reason, today actually we released our FAL agent. It's our creative harness to make it easy to work with these video models.

And we are very excited how people are going to use it.

Sophia (4:15)

Oh, I want to unpack that a little bit more. Before we go into that, wanted to talk more about token content. Token market fit. Token market fit. I love that term.

So my understanding of it is, all right, the number, the amount that we're spending on token costs is now worth it. Because if you were to do this a year ago, even a few years ago, you spend an arm and a leg to generate a 30 second, three second clip and it's awful. It's not even that great. So sure, maybe it was possible, but the amount that you would spend just wasn't worth it. You could maybe generate it for a way cheaper.