Anastasis Germanidis, Runway

Clips
News
More from Ray Summit 2024

Anastasis Germanidis

Co-Founder & CTO

Runway

Three insights you might have missed from Ray Summit

The rapid rise of generative artificial intelligence is driving significant changes in the enterprise landscape, and companies such as Anyscale Inc. are at the forefront of this transformation. With the complexity of AI infrastructure increasing, businesses are turning to innovative solutions to help them scale AI efficiently

AI-driven filmmaking: How Runway is transforming visual storytelling and creativity with Gen-3 Alpha

As artificial intelligence continues to take center stage in the digital landscape, AI-driven filmmaking is becoming increasingly important for enhanced audience engagement, innovation and creativity. Runway AI Inc. heeds to this call since it uses generative and novel storytelling to produce compelling visuals and personalized content, according to Anastasis Germanidis (pictured), co-founder and chief technology officer of Runway. Runway’s Anastasis Germanidis talks to theCUBE about the revolutionary effect of AI-driven filmmaking.“Madonna’s live concert tour used Runway for all the visuals that were being displayed during her performances,” Germanidis stated

play_circle_outline Runway used for Madonna's live concert visuals, enabling novel and compelling storytelling.

play_circle_outline Runway empowering filmmakers to experiment with new ideas and reduce production costs.

play_circle_outline Runway Studio funding projects, collaborating with Lionsgate, and promoting generative media community.

play_circle_outline Versatility and developer experience of tools like Ray crucial for building video models.

play_circle_outline Continuous surprise and discovery in building new models for coherent stories and narratives.

Info
Transcript

Anastasis Germanidis, Runway

Anastasis Germanidis

Co-Founder & CTO Runway

Anastasis shares the latest updates on Runway's tools, including the video-to-video tool, which has been well-received in the creative community. Runway's collaboration with Madonna's live concert tour demonstrates the tool's versatility and creativity. The Hundred Film Fund and the 48-hour film festival also showcase the potential of AI-driven filmmaking projects. Runway aims to empower both new and established filmmakers to experiment with ideas and lower the barrier to traditional production costs. The partnership with Lionsgate further emphasizes Runway's... Read more

explore Keep Exploring

What was one unexpected use case for Runway mentioned recently, involving visuals for Madonna's live concert tour? add

What is the role of Runway in bridging the gap between AI/technology and art/creative tools, and how does the creative team work to help artists incorporate Runway into their projects? add

What is the purpose of Runway Studios and how do they plan to support creators and artists in their projects? add

What is the importance of versatility in model training, particularly in regards to video models compared to language models? add

What is the approach to planning and discovery at Runway and what is the goal when building new models? add

bolt Powered by CUBE AI

Anastasis Germanidis, Runway

Savannah Peterson

>> Good afternoon, AI community and welcome to Ray Summit by Anyscale. We're here on our home turf in San Francisco, California. My name is Savannah Peterson, delighted to be broadcasting live all afternoon with some of the fantastic guests and keynote speakers that have been on this stage. We are starting today off with such a bang. You may be wondering about AI filmmaking. If you are, you've certainly heard of Runway. Anastasis, thank you so much for taking the time to hang out with me today.

Anastasis Germanidis

>> Of course. Happy to be here.

Savannah Peterson

>> How was your keynote this morning? I got to see it packed room, standing room only. It seems like you've got to be getting used to telling people about AI filmmaking. Were you at all nervous or excited about today in particular given how big this showroom floor is?

Anastasis Germanidis

>> I think the interesting part is that just because the field is moving so fast, our tools are evolving so fast, every time I get to talk to an audience, we have something new out, so it's something different every time. So that's exciting. I think it's great to talk about some of the work that we've been doing on the research side and on our tool side and give a glimpse of what's also coming out.

Savannah Peterson

>> You did a good job of giving us a teaser and we'll get into that in a second. What are some of the tools that you've adopted since the last time you were able to present? What are some of the newest tools in your kit?

Anastasis Germanidis

>> So most recently we released our video-to-video tool. So that allows you to take an input video and translate it while maintaining the structure, completely transform the style of that video. So that's a really versatile tool because you can really control the exact motion that you want to generate while at the same time produce really, really amazing stylistic results with it. So we see people using that kind of complementary to image-to-video and text-to-video, a place for all those different tools.

Savannah Peterson

>> So that would make sense as the evolution. And when it comes to our visual language, it's not just language, it's visuals, there's context. As you said, as you were saying in your keynote, I loved the controllability aspect of what Runway does is it allows you to take people into a new portal or a new dimension or do things that have never been done before. Do you have any examples from recent times where you were really surprised by something that someone created or something that was very visually intriguing you don't think we could have made before?

Anastasis Germanidis

>> Yeah, so one example that I happened fairly recently was Madonna's live concert tour used Runway for all the visuals that were being displayed during her performances.

Savannah Peterson

>> Wow.

Anastasis Germanidis

>> So this is a use case that was not necessarily what we originally considered Runway for, but it was a really nice example of using the visuals and the generative storytelling to produce really compelling visuals. And we saw talking to the artists that created those visuals, they specifically said that they were able to really get away from the more traditional VFX look and create something that feels novel and that feels different than what's already out there. And so this is exactly what we envisioned Runway to be able to create.

Savannah Peterson

>> I was going to imagine that had to be music to your ears. I just have to know, when you co-founded Runway, were you expecting the call from Madonna's team to use your product to show off art to millions of people?

Anastasis Germanidis

>> Well, the great thing is very often those tools are being used without even us knowing. People find out by themselves that they can use Runway for really interesting use case and they just adopt the tool. So that's what we want to see. It's not necessarily us reaching out to specific artists and trying to get them our tools. We want people to really find Runway to be useful and find Runway to be a creative extension of themselves.

Savannah Peterson

>> Absolutely, and I think that becomes a bit of a controversial topic when people think about art and AI. And I know what side of the fence that we sit on, but I think for folks still learning about how these tools are going to augment that creative process, this is a really great example. I mean, especially, I don't know if Runway was used when she did her concert for 1.6 million people or whatever it was down in Rio, but you can't see someone with your eyes if there's 1.6 million people in the crowd. So you've got to have cool visuals that are an extension of that expression and not just video. What are some of the other things that you're excited to see people creating? I know you have your AI Film Fest this year. You had five extra submissions of last year. Congratulations. Very exciting. Anything interesting come out of that?

Anastasis Germanidis

>> So yeah, a lot of amazing films on the festival. Every year it feels we reach a completely different level of quality of consistency with those movies that are being made. We also recently announced the Hundred Film Fund and the goal of that is to more directly support specific AI driven filmmaking projects. And we also are doing the 48-hour film festival. So there was a few weeks ago, you have essentially a weekend to create a film and we've done that three times. And this time was the first time people could use Gen-3 Alpha and it was on a completely different level of the kinds of stories that were being told, the kind of stylistic diversity of the videos, the kind of narrative and experimental aspect. It was really amazing to see what people created and that's only going to continue as the models get better, as more people find out about Runway. And I think that's at the core of how that conversation, I think, will evolve. Is the moment you actually use the tools by yourself, you realize it becomes demystified and you see them as just a next technological leap for our tools. It's not necessarily some entirely new, it's not like that the AI is creating the artwork by itself, which is a common perception about those AI-driven tools. It's really creative giving you superpowers allow you to make more choice in a shorter period of time. And I think what's really exciting about those tools.

Savannah Peterson

>> It is. And you can agree or disagree with this or share your perspective, but when I think about art or creating anything or even what we're doing right now, I feel like sometimes what we're trying to do is press print on our brain. You're trying to communicate some vision that you can see in your head that doesn't necessarily translate to a traditional storyboard or to sketching or even the ability to verbally communicate sometimes. And I think, and I'm curious to what you think, I think that AI as a creative tool is helping us unlock more of that, almost like a Z-axis within the creative mind that was maybe stuck a little bit by the tools that we had beforehand or at least the skills required to use those tools.

Anastasis Germanidis

>> Yeah, exactly. We see it as a new camera. It's all about where you point-

Savannah Peterson

>> Oh, it's a great way to put it.

Anastasis Germanidis

>> You point it to worlds that are not yet exist that are generated as you point the camera, but ultimately, it's really about your agency as an artist and your ability to have a very specific vision you want tell. So we see the same thing that applies to previous tools applies here. It has a learning curve, it pays off to really spend a lot of time with the tool. And we see some people becoming really exceptional with the tools, similar to people achieving mastery in previous generation of tools. So nothing is different from that regard. It's just that the tools hopefully empower people to create more in a short period of time.

Savannah Peterson

>> Yeah. Do you think that this is empowering or even as some might say, democratizing the creative process with tools like Runway?

Anastasis Germanidis

>> I think so. And we see that it's not just new folks that are now entering this filmmaking world or creating artwork for the first time. A lot of established filmmakers are finding value from those tools. Something that we often hear from especially filmmakers is they can use Runway to experiment with ideas that they previously was not sure if it was an idea worth investing in.

Savannah Peterson

>> Right. Because it is very expensive. Traditional production can be quite expensive, especially with VFX.

Anastasis Germanidis

>> Yeah. So that's lowering the barrier to actually trying out an idea and will allow people to really just see, oh, is this something that...? If something feels too risky of an idea to experiment with, now you can actually visualize it for the first time and then you get a better intuition about is this something that I need to pursue more and build a larger story around? But you reduce that hesitation of should I actually pursue this, because now you can actually just quickly try it on your own and then if it's worth investing more on, you can do that.

Savannah Peterson

>> Yeah. Well I think you brought up a really great point when we were chatting before we went live about the development process within film. It's not just about a two-hour feature film, it's also about getting that buy-in, it's about pitching, it's about selling that MVP or even getting funding from your Hundred Film Fund for example, as you want to be able to show it off. It seems like there's a lot of opportunity for creators to pump this out to give a little teaser. Do you see that happening a lot?

Anastasis Germanidis

>> Yeah, so the way we think about the adoption of tools like Runway is it doesn't have to be a thing where you go from zero to not using generative models at all to immediately using it for every your project, it can be very gradual. You might want to use Runway for, be it on the pre-production side. You want to do some storyboarding, just test out the words, see how the shape of your story looks like. And if you have a good experience with it and you find value in it, and you can potentially use some shows that are generating Runway in your actual output and your actual film. But it's really whatever part of the creative workflow, Runway can help in. I think it's fine. It doesn't have to be zero to one.

Savannah Peterson

>> You're happy to just be a part of the process. It doesn't matter which piece of that process necessarily.

Anastasis Germanidis

>> Right.

Savannah Peterson

>> Yeah. Would you say, so I've got some wonderful friends in Hollywood, shout out to all my friends who know who you are, that are a little nervous about AI. What would be your advice to them? How could they get started or try it out? How do you hand hold people who are coming from the older side of film production?

Anastasis Germanidis

>> So a lot of the way we see our role is to essentially become translators between those two very different languages. The language of AI and technology and the language of art and creative tools. And so that's really our role is we have a creative team working with artists every day trying to get them to really working to understand how to incorporate Runway and if Runway is useful in their own projects. So there is definitely some hand holding involved and I think we see that as part of our role is to educate and show what you can do with those tools. But I think the best way to start is really, Runway's available for everyone to try. You can go in and try it out. We do invest a lot on building more partnerships with entertainment industry. We announced a partnership with the Lionsgate recently.

Savannah Peterson

>> I was just about to bring that up. Tell us about that.

Anastasis Germanidis

>> Yeah. So again, very excited to work with film studios directly and work with the best storytellers to see how generative media, generative tools can be used in their workflow. And Lionsgate, their masters of their craft. So we're very excited to collaborate together. We're starting from training a custom model on Lionsgate Gallery, but also we just generally are trying to figure out what are different projects, what are different, where Runway can really provide value and we can build specialized tools to help with that specific project. I think there's still a lot of discovery and a lot of things that we need to figure out in order to make those models useful. And for us, we see our role now is building the models but actually really crossing that cousin between cool results and cool outputs and something that feels like an extension of your creativity and extension revision.

Savannah Peterson

>> Yeah, it's different to say, wow, it's another thing to be completely moved by say a feature film. So how long are we away from a category at the Oscars for a Gen-AI feature film?

Anastasis Germanidis

>> Well, hopefully it's not a category. Hopefully again, Gen-AI becomes demystified and those tools just become yet another part of the filmmakers' tool books. So we actually don't want a separate category-

Savannah Peterson

>> I like, this is good, this is good.

Anastasis Germanidis

>> We think that if those models are useful, they're going to be part of films that are being made in every category-

Savannah Peterson

>> So it's going to be like when we got better lighter cameras and a lot of other things that happen on the editing side. I like it. Okay.

Anastasis Germanidis

>> The moment we're successful is when people don't talk about the generative models that might be used in a film. It's like they're using the best tool for the job.

Savannah Peterson

>> Yeah. Yeah. So what is next? You mentioned the Hundred Film Fund. Who's funding those hundred films? Is that you as a company? Are there partners?

Anastasis Germanidis

>> Yeah, so we have a part of Runway called the Runway Studios that's working with creators and artists. And teams of creators and Runway Studios is funding this effort. And the idea is to support projects that we find really compelling, really promising, give the time and space for those artists to really develop their project. And it could take a different shapes and forms. We don't have a specific idea in mind of what stories could be told, we want to be surprised. But we do want to find more ways of supporting the generative media community, figuring out what's next for... And that will hopefully also inform the tools that we build and how we develop product moving forward.

Savannah Peterson

>> Yeah, absolutely. Well, I can't wait to see those hundred films that you end up funny. It's going to be-

Anastasis Germanidis

>> .

Savannah Peterson

>> It's going to be really interesting. I think we'll all be surprised and delighted. We're here at Ray's Summit. We've talked a lot about tools. How important is it for you to have tools like Ray that can handle the variety of different models and things that you're constantly working with?

Anastasis Germanidis

>> I think versatility is really important. As I mentioned in the presentation, there's less standards for video model training as there's for language model training. And we're essentially inventing the paradigm as we go, as we build those models, as we scale them even further. And what's nice about Ray is that it has a really nice developer experience and you can build those new components and those primitives. And so we use Ray on different aspects of our . On the data and training pipeline in order to accelerate our efforts to build those models.

Savannah Peterson

>> It makes a lot of sense. When you were speaking at the keynote, I was struck by realizing how much we don't understand our visual environment and building models around that. What's the biggest challenge or difference between building models that are based on our visual world versus our language?

Anastasis Germanidis

>> So there is specific engineering and technical challenges involved with working with video versus with language. First of all, in terms of size, you have it's order of magnitude larger data sets and you do-

Savannah Peterson

>> So much data.

Anastasis Germanidis

>> With language and it can also take maybe different shapes and forms, maybe different resolutions, aspect ratios. So there is different aspects that you need to deal with specifically when you're working with visual data. I think the other side of this is that there is a lot of subjective evaluation that must go on when you're building those models. You can-

Savannah Peterson

>> I was wondering about that. Yeah, dig in there.

Anastasis Germanidis

>> You can try to develop quantitative metrics to evaluate different aspects of the model's, understanding of different concepts, of the most understanding of camera movements, the ability to render text, all those different capabilities. But ultimately, you're making visual content that's meant to evoke feelings, to have people have visceral reaction to that.

Savannah Peterson

>> Exactly. The emotion of it.

Anastasis Germanidis

>> And that's hard to capture in one specific metric. So we're working together with a creative team, we're doing a lot of more subjective evaluation, more paying attention to the aesthetics of the outputs of those models, to really trying to make sure that the outputs of those models are able to be used in creative purposes.

Savannah Peterson

>> Yeah, it's a lot of different factors to be thinking about within even a single piece of visual data. Oh, super exciting. All right, last question for you. You talked about how everything's changing all the time. You're changing tools, there's all this exciting stuff going on with your film fest and the fund. What do you hope to be able to say the next time we sit down at Ray's Summit, for example, a year from now that you can't yet say today?

Anastasis Germanidis

>> It's a great question. A big part of how we think about planning at Runway is discovery and the ability to be surprised at the very core of what we're working on. And if I could predict fully what I'm going to be saying a year from now, we maybe would not have succeeded in our goal because with every new model that we build, we're always surprised at the capabilities that are present in that model versus the previous generation models. So we do expect that you're going to be able to tell much more coherent stories where you can have character consistency, environment consistency. You're going to be able to have scenes with dialogue and more conflict interactions than we have today. That's fairly straightforward to predict, but in terms of the actual stories and the actual narratives and movies that are going made with those models, I really hope we're surprised by gets created. I think that's the fun part of building those tools. It's like that people constantly surprise us with what they create.

Savannah Peterson

>> I love that. So actually your answer to that question is not so much what you hope to be able to say, but what you hope not to be able to say, which is we expected this, that we're consistently surprised and . Thank you so much for taking the time with us on what is undoubtedly a very busy day. This has been delightfully educational. And thank all of you for tuning in to our wonderful full day of coverage here at Ray Summit by Anyscale in San Francisco. My name's Savannah Peterson. You're watching theCUBE, the leading source for AI tech news.