Models, Modalities, and Memes: Creating Compelling AI Characters

**Michael Lingelbach** (0:00)
It was really interesting, as you see this crossover between consumer and enterprise use cases of AI, where consumers will go out and they'll do the hard work to figure out, oh, I can make a Talking Baby podcast, and then I'm on Forbes the other day, and I see a Hedra generated ad, and I text Justine, and I'm like, what? How did this happen?

**Justine Moore** (0:18)
With the Talking Baby advertising some enterprise software or something.

**Michael Lingelbach** (0:21)
Yeah, and it was weird, because it was like, first of all, the Talking Baby stuff, that was so last month. People were like, I was kidding. But I mean, it's just interesting that now we're in the era where companies are moving really fast now. Like they're really getting excited about adopting AI. And I think it's our job as startups to kind of take these signals from the prosumer users and then productionize that and make that something that enterprise users can come and rely on to make the next generation their content needs.

**Eric Thornburg** (0:50)
Thanks for listening to the a16z AI podcast. I'm Eric Thornburg, general partner at a16z. If you've been following the rise of generative video, you've probably seen some of the wild outputs. Talking babies, podcasting anime characters, and even AI newscasters. But behind the scenes, a different kind of innovation has been unfolding. One that's less about spectacle and more about control, speed, and character performance. In this episode, Hedra CEO and co-founder, Michael Lingelbach, joins a16z partners, Justine Moore and Matt Bornstein to talk about how Hedra became the engine behind millions of these viral clips and what it takes to build video models that feel truly alive. From vertical integration to interface design, from prosumer creativity to enterprise scale, this is a conversation about much more than rendering pixels. You'll also hear about how the Hedra team is thinking about real-time interactivity, programmable personas, and the UX of what they call character-first video. So stick around and hear Michael, Justine, and Matt dive into the future of generative media right after these disclosures.
As a reminder, please note that the content here is for informational purposes only. It should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security. It is not directed at any investors or potential investors at any a16z fund. For more details, please see a16z.com/disclosures.

**Michael Lingelbach** (2:12)
You may have seen some of the viral content that was produced with us in the past couple months, ranging from everything from those fun anime-style characters a couple months ago, to the baby podcast, to... I actually don't know what's trending this week. I think memes are one of the best go-to-market strategies for a lot of companies. It's kind of crazy the way that you can gain mind-chair just by getting in front of so many people. I think this is becoming a really common strategy with a lot of startups. I mean, there's another a16z startup, Cluely, that definitely has leveraged a lot of the virality of Twitter to that recognition. I think, though, what memes really are is that this technology has allowed people to be very creative very quickly. And I think memes are like the manifestation of that creativity. Is this short-form content has really come to dominate the cultural consciousness? And one of the things that Generated Video does is it lets you take any idea and make it into that content in a matter of a couple seconds.

**Matt Bornstein** (3:04)
So explain a little bit why people use Hedra for memes and how they do it, and maybe how that connects to some of the markets that you're really going after.

**Michael Lingelbach** (3:12)
Yeah, I mean, we were, I think, the first major audio-video foundational model company that was able to deploy these full-body, expressive, dialogue-centric Generated Videos at scale. And people create millions upon millions of pieces of content with us now. And I think the reason that we really took off was because this was kind of a core missing element of the stack of how people were creating content. So things like creating Generated Podcasters, creating animated characters, talking in scenes, creating singing. It was stuff that was really challenging to do before us. It was either too expensive, it wasn't flexible enough, because we were the first to really bring this image to audio-video model technology to market, or it was too time-consuming. Our model was very quick, it was very cheap. And I think as a consequence, we've seen the rise of these virtual influencers. CNBC had a really good article on virtual influencers powered by Hedra recently.