How Real-Time AI Video Is Changing How Creators Work

Watch on YouTube ↗  |  September 17, 2026 at 17:33  |  38:57  |  a16z
Speakers
Gorkem Yurtseven — Co-founder, fal
Batuhan Taskaya — Head of Engineering, fal

Summary

a16z General Partner Jennifer Li interviews fal co-founder Gorkem Yurtseven and Head of Engineering Batuhan Taskaya about fal's H3 Max, a post-trained version of MiniMax's open-weight video model. They explain how model post-training, system optimizations, and hardware advances reduced generation time by an order of magnitude, enabling real-time and continuous AI video. The conversation covers technical memory, cost economics, consumer live-streaming experiments, and the push for controllability to unlock Hollywood and professional creative workflows. Main market implications include strong demand for generative media, continued compute constraints and demand for advanced AI chips, and rapid AI adoption by Hollywood studios.

  • fal post-trained MiniMax's open-weight H3 video model to create H3 Max.
  • Post-training and system optimizations cut generation time by about 35x while maintaining quality.
  • H3 Max Turbo can generate a 5-second video in about 1.5 seconds at 2x lower cost.
  • Real-time continuous video streams with memory and action control went viral after launch.
  • Latest hardware like Blackwell/GB200 is needed for beyond-real-time video generation.
  • fal sees Hollywood as its fastest-growing segment and expects studios to massively increase AI usage.
  • The next focus is controllability: camera, lighting, lip sync, motion, and character consistency.
Ideas
Gorkem Yurtseven Co-founder, fal 0:00
Generative media demand strong, compute-constrained.
Generative media has reached token market fit alongside coding agents, with individual professionals willing to spend thousands of dollars per month on video generation. Demand is so strong that the industry has been compute-constrained since April, and post-training plus system optimizations that make models an order of magnitude more efficient should relieve that bottleneck and enable more usage, supporting the generative media theme.
Gorkem Yurtseven Co-founder, fal 0:24
Hollywood studios to massively increase AI usage.
Hollywood is fal's fastest-growing segment; AI usage was nonexistent a year ago and studios now need controllable point solutions for camera, lighting, lip sync, and motion rather than fully generated video. With post-training infrastructure and US-hosted models addressing legal and data-residency issues, studios are expected to 10x-100x their AI usage, making Hollywood a key beneficiary of AI video tooling.
Batuhan Taskaya Head of Engineering, fal 9:33
Advanced AI chips needed for real-time video.
Latest-generation AI hardware such as Blackwell and GB200 provides a 2-3x speedup over Hopper and is necessary to generate video faster than real time, though cost is comparable. This indicates continued demand for advanced AI semiconductors to support real-time video generation workloads.
Up Next

This a16z video, published September 17, 2026, features Gorkem Yurtseven, Batuhan Taskaya discussing Generative media, Hollywood studios, SMH. 3 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Gorkem Yurtseven, Batuhan Taskaya  · Tickers: Generative media, Hollywood studios, SMH