AI + a16z

Models, Modalities, and Memes: Creating Compelling AI Characters

Episode Summary

How do you build real-time, expressive AI video from first principles? Hedra CEO Michael Lingelbach joins a16z to discuss model architecture, UX tradeoffs, and why character-centric design unlocks the next wave of generative media.

Episode Notes

Models, Modalities, and Memes: Creating Compelling AI Characters
In this episode of AI + a16z, Hedra founder and CEO Michael Lingelbach joins a16z partners Justine Moore and Matt Bornstein to talk about building AI-native video — and why the next wave of generative content is all about characters, not just clips.
They discuss how Hedra’s expressive, full-body, dialogue-centric video models are powering everything from viral meme content to enterprise training tools. Michael explains why “character” is the core design primitive in Hedra’s architecture, how consumers are leading the charge in discovering new use cases, and what it takes to productionize those behaviors for real-world applications.
Along the way, they explore what makes multi-modal generation uniquely hard, the role of user control in shaping believable AI performances, and why being a founder sometimes means responding to thousands of support emails — at 6 a.m.

Key takeaways:

• How Hedra’s real-time video model blends audio, image, and character control
• Why generative content is shifting from static avatars to programmable personas
• The surprising crossover between consumer creativity and enterprise adoption
• Where existing LLMs fall short in generating emotionally authentic characters
• What vibe coding, hands-on design, and founder obsession look like in practice

For anyone curious about building AI characters, scaling creative workflows, or the future of human-computer interaction — this one’s not to be missed.