We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode Runway Has AI Video Breakthrough "Act One" for Face Rigging

Runway Has AI Video Breakthrough "Act One" for Face Rigging

2025/2/11
logo of podcast Lex Fridman Podcast of AI

Lex Fridman Podcast of AI

AI Deep Dive AI Chapters Transcript
People
播客主讲人
Topics
播客主讲人: Runway最新推出的Act One功能为AI视频制作带来了突破,它能将演员的表演实时转化为不同的角色,并同步口型和表情。这项技术极大地简化了动画制作流程,降低了创作门槛,使每个人都有机会成为制片人。我看到了Act One的巨大潜力,它将推动AI视频技术的发展,并在2025年迎来AI视频的爆发年。然而,我也意识到这项技术可能被滥用,特别是深度伪造的风险。为了应对这些风险,Runway采取了一系列安全措施,包括内容审核、公众人物检测和声音版权验证。虽然这些措施并不完美,但表明了Runway对安全问题的重视和责任感。我希望Runway能继续完善安全机制,确保Act One在推动创作的同时,也能最大限度地减少潜在的负面影响。我将持续关注Act One的进展和应用,并及时分享最新的信息和见解。

Deep Dive

Shownotes Transcript

Translations:
中文

Runway, the creator of AI Video Generator, has just come out with an amazing new feature they've just launched called Act One.

There's a ton of people talking about this. I want to break down a couple of the use cases, a couple of the demos that people are showing off and what exactly this tool can do. Now, before I get into it, I wanted to mention that if you are looking for a good way to make money with AI tools or grow your current business and scale with AI tools, I would love for you to be a member of the AI hustle school community. This is something where I put exclusive content every single week.

of all the tools, workflows, and products I'm selling, the ways that I'm making money with AI tools and how I'm growing my businesses. So if you're interested in joining the community, I would love for you to be a member. It's currently $19 a month. I'll probably increase the price to $100 a month in the future. But for now, if you lock in that price, it will never be raised on you.

Okay, let's talk about what Runway is currently working on. So Act One is absolutely fascinating. They put out a blog post, they announced it over on LinkedIn and on X recently.

And they've showed off some incredible demos. Essentially what this does is it allows you to have one actor sitting in front of, you could say a green screen, but really like a white screen and they're talking and it converts it into some sort of character, realistic or cartoon. It's using the same words as using the same expressions. The head is moving the same way, but it's a completely different character and you can do a completely different voice that syncs to the lips.

All in all, this is absolutely incredible. They have a ton of cool demos, I want to show you a couple interesting things that people have made with this, how this works. But I wanted to say show you a little bit of the responses first on in the comment section on x what people are saying. People are saying the motion capture industry is cooked. This is true. This is a massive industry. People are saying how long before Chinese overlords copy this with less censorship,

better UI. Give it three weeks on a side. Now this is pretty great. I have to subscribe again. Okay. So, you know, some people are saying Chinese overlords, blah, blah, blah, whatever. I think like the real thing people are concerned about is deep fakes are now theoretically incredibly easy, right? You could say anything and now it can turn into me being Trump or Joe Biden or, um, you know, anyone else, Xi Jinping, any, you know, any leader or whatever celebrity. And obviously this could be very problematic. Now,

Runway has specifically addressed this concern as they kind of talked about the rollout of Act 1 and what they're doing for safety. They said with...

As with all our releases, we're committed to responsible development and deployment. We're releasing this new tool with a comprehensive suite of content moderation and safety precautions, including apparently they have even more robust abilities to detect and block attempts to generate content containing public figures, right? So, and it's kind of interesting, like what,

constitutes a public figure, obviously actors, obviously government officials. It's interesting to see if it goes much further than that. Is it YouTubers? Is it, you know, whatever. So they say technical measures to verify that an end user has the right to use the voice they've created. This is also interesting, right? Because you're technically people could be trying to use it for deep fake

and then just cut out the video. So make it sound like Joe Biden or Trump or anyone else, and then cut out the video. And then you've got that. Now there's tons of other tools that actually technically allow you to do that. I'm pretty sure. Um,

There's lots of open source tools where you can clone voices. And you could do a lot of that. So I'm not saying that this is completely impossible. The video aspect, though, is quite a big step forward. And it's different. So they're really trying to counter that. They also said continuous monitoring to detect and mitigate other potential misuses of our tool and platform. Okay, this doesn't mean anything continuous monitoring. I mean, it's just saying, if something goes viral on Twitter, they're gonna quickly try to fix the bug of whatever bad thing someone used it to generate. So I mean, that's

Every company is probably doing that or should be doing that. So I wouldn't give them a ton of credit for that. That being said, like, I do appreciate they have this whole safety blurb on here. And this is something that they're actually looking forward to. They said, looking ahead, we're excited to see what forms of creative storytelling Act One brings to animation and character performance. Act One is another step forward in our global and our goal to bring previously sophisticated techniques to a broad range of creators and artists.

We look forward to seeing how artists and storytellers will use Act One to bring their visions into life in new, exciting ways. Access to Act One will roll out to everyone soon. Okay, so what's interesting to me is they're like, we're trying to see what people are going to use this for. They have an example of

one guy, they have this like kind of cool skit or shot or whatever, where it's one guy playing both roles of two people in a diner sitting across from each other talking. And he's just in front of a green screen reciting both lines, like he looks one way recites a line changes his head looks the other way recites the line. And they use that to do both characters completely different voices, completely different sets, completely different characters,

This is amazing. This is, you know, stuff that Hollywood has been doing for a long time and kind of had the capabilities of doing this to some degree. Um, this is completely a whole new level and this puts it into the hands of literally everyone. So really, really interesting. And I mean, even to the point where people are like, Oh no, what about like deep fakes? Like obviously the capability has been there. Hollywood's been doing deep fakes. You can do makeup. You could do, there's all sorts of things you could try to, you know, clone someone or whatever. Um,

This is now just in the hands of everyone, everyday people. So very, very fascinating. Something else that I thought was really interesting is they, you essentially put in like the driving performance. So the original person, it's not even a green screen, they put it on a white screen. So just a plain background. And they've shown a couple of demos where they have the main driving performance. They're doing a ton of different shots. There's like one guy and then he's, you know,

very clean, um, sort of animated looking in front of a bookshop. There's another one where it's like the same dude, but he's like, uh, very kind of overexposed background. He's in a car. Um, and another one where he's got like a ton of hair and he's, you know, in, in a white room. So it's the same guy in three different settings. What was interesting to me is that his face is actually looking in a slightly different direction in the three settings. Um,

Overall, everything Runway has produced here is absolutely incredible. I'll definitely keep you up to date on anything else they pull out and also what people are making with this. I think we're about to start seeing on Twitter and X a ton of different content people sharing where they've actually used these tools, you know, recording themselves and making multiple people to make some incredible films. This is next year. 2025 is the year of AI video. I believe everyone's going to become or have the tools to become a

producer and create their own films. I'll be covering all of that and keeping you up to date. If you enjoyed the podcast today, if you could leave it a review, I'd be super grateful. And if you're interested in making money with AI tools, make sure to go and check out the AI Hustle School community, link in the description. Thanks so much, and I will see you next time.