Posted May 15, 2025

There is something special about visual storytelling. Nothing strikes an emotional chord quite like the dramatic conclusion to a great film. Nothing helps us understand current events like seeing it happen for ourselves, or grabs our attention — for better or worse — like the TikTok feed. Even corporate training content has higher recall in visual formats.

So, video is an important yardstick to measure progress in artificial intelligence. And recent progress has been unbelievably good. Just two years after the original spaghetti test, it’s possible to generate photorealistic video clips of people and environments using only AI. There’s a burgeoning community of creators making everything from funny clips to Hollywood-grade short films.

There’s only one problem: the “actors” in AI videos can’t really do much. Most AI tools struggle to produce natural, expressive characters — particularly if said characters need to speak. The historical standard is applying a voice to a static character image with lip sync, but this isn’t always convincing and takes the viewer out of the story. It’s the return of the uncanny valley.

That’s why we’re thrilled to announce our Series A investment in Hedra, one of the leading companies working on this problem.

Don't want to read any further? Watch Justine and Matt read this post via a Hedra video.

Hedra is an AI video company focused on character animation, making it easy to generate a character-driven video scene from an image and audio track or text script. Hedra supports any input image (not just a pre-built character) and generates realistic lip syncing, facial emotions/gestures, and background animations. While general AI video companies also offer some of these features, we believe Hedra’s outputs are distinctly coherent, engaging, and natural.

This technology is far from competing with human actors. And the quality and control of the outputs are not nearly sufficient for feature film releases. We certainly don’t think traditional filmmaking will ever go away. But AI systems like Hedra open up visual storytelling to a much wider audience than ever before, and, like CGI, they give serious narrative filmmakers a tool to try out new ideas and expand the range of stories they can tell.

So far, nearly 3 million users — from consumer meme makers to enterprise marketers — have generated over 10 million videos on Hedra. If you’ve been on social media over the past few months, you’ve probably seen a viral video created with Hedra (e.g., podcasts, trailers, title sequences, streamers, and ASMR videos). We believe adoption of the platform will only grow as the company expands the control, range of motion, and expressiveness of the model.

Hedra participated in the a16z speedrun program, so we’ve been privileged to get to know Michael Lingelbach and the team over a period of several years. We’ve been incredibly impressed by their velocity in shipping state-of-the-art models and features. They’re relentlessly focused on pushing the standard of AI-powered storytelling, and they have a unique combination of deep technical and research expertise with clear product thinking.

If you’re looking to create a talking character, we’d highly recommend giving Hedra a try if you haven’t already. We think you’ll be impressed by what you find.