a16z Podcast: Real-Time AI Video, Controllability, and Hollywood’s New Workflow

**SPEAKER_1** (0:01)
Powerful AI video is moving from novelty to production tool, and this episode of the a16z Podcast shows exactly how. We're diving into H3 Max, a new video model that Gorkem Yurtseven and Batuhan Taskaya built around speed, quality, and what they call token market fit. And their message is really striking—it's not just about better video anymore. It's about creating a workflow that creators, studios, and technologists can actually use every day. So what's making this different? Performance is a huge part of it. The team says H3 Max Turbo can generate a 5-second video in about 1.5 seconds, at half the cost. We're talking about a 35x speedup over the original Minimax without sacrificing quality. That's incredible. And they're framing this as a system-wide win, right? Post-training, kernels, hardware optimization, all pushing utilization far higher than before. The goal is to move beyond the usual limits and make video generation dramatically more efficient. Exactly. But just as important, H3 Max is open-weight and open-source, which the speakers describe as a major shift. That openness helped spark viral attention, but it also created an ecosystem. What does that ecosystem look like in practice? Creators are now combining GPT, Blender, and H3 Max to build beautiful scenes, try ideas in parallel, and accelerate professional production. For Hollywood and VFX teams, that means a new workflow built around iteration, references, and control rather than starting from scratch.
That word controllability comes up again and again. They argue that the biggest gap in the market is controllability. What kinds of controls are we talking about?
Text-to-video, image-to-video, reference-to-video, camera controls, lighting controls, motion transfer, and lip-syncing. These are the next frontier. One especially ambitious system, H3 Max Director, can generate up to 60 minutes of continuous, action-controlled video.
Sixty minutes? That's remarkable. And it remembers recent scenes, so the conversation with the model feels more like directing a film set in real time. The impact must be significant.
It's already visible. The model has become the most popular on their platform by nearly double the volume, and Hollywood is now their fastest-growing segment. Studios want point solutions that extend footage, change angles, adjust lighting, and fit specific production needs.
What about the practical barriers? I imagine legal concerns are a big issue for major studios.
They are. But the speakers note progress on legal approval and US-hosted instances, which they say removes that final barrier for major studios. Looking ahead, they believe these tools could drive a 10x or 100x increase in AI usage across the industry. From novelty to essential production tool—that's quite a transformation.