AI Summary
5 min readWhen Video Generation Becomes Real-Time
FAL’s H3 Max model can generate five seconds of video in 1.5 seconds at one-twelfth the cost of the original Minimax H3 model—and it emerged from a weekend project that nobody at the company had planned. One engineer started live-streaming continuous generations from his laptop on Twitch. Another team built an experience called FAL Live that let viewers vote on what happened next in a scene. Within days, multiple unplanned products were going viral simultaneously. The speed wasn't just an incremental improvement; it fundamentally changed what was possible.
The Compounding of System-Level Optimizations
The dramatic speed and cost improvements didn't come from a single breakthrough. FAL’s head of engineering, Batuhan Taşkaya, describes it as a compounding effect across multiple optimization variables. The most important was post-training the model to run in fewer diffusion steps—from 50 steps down to about 20—without sacrificing quality. This alone gave an order-of-magnitude speedup.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of a16z Show
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:00) **Generative Media's "Token Market Fit" Moment** - The hosts introduce the episode's core premise: that generative video has reached an inflection point where speed and cost make large-scale consumer use viable.
- 2 (01:39) **Introducing H3 Max: A League of Its Own** - Gorka and Botan describe the model's launch and its surprising performance against existing benchmarks.
- 3 (03:21) **The Strategy: Post-Training Open-Weight Models** - The team explains why a general inference platform like Fal would invest in post-training an open-weight model like H3.
- 4 (06:20) **The Secret Sauce: Compounding Optimizations** - Botan details the multi-layered approach that produced a 35x speedup without sacrificing quality.
- 5 (09:55) **Hardware and the Path to Real-Time** - The discussion covers current serving configurations and the role of next-generation chips like Blackwell.
- 6 (10:42) **Beyond Turbo: The "Max" vs. "Turbo" Trade-off** - The team reveals they created an even faster version called H3 Max Turbo that generates a 5-second video in 1.5 seconds at 12x lower cost.
- 7 (12:56) **The Surprise of Real-Time and Internal Spontaneity** - Gorka recounts how the model's speed was so surprising that the team waited for external validation before announcing it, and how it sparked a wave of unsanctioned, creative projects inside the company.
+ Full timestamped outline available in the app
Show Notes
a16z General Partner Jennifer Li sits down with fal co-founder Gorkem Yurtseven and Head of Engineering Batuhan Taskaya to discuss what changes when generative video becomes fast enough to run in real time.
They unpack the technical work behind H3 Max, fal’s post-trained version of MiniMax’s open-weight video model, and how combining model post-training with systems and hardware optimization significantly reduced generation time while maintaining quality. That speed has enabled experiments with continuous video, including streams that can remember previous scenes and respond to new directions while they’re running.
They also discuss why the next challenge may be less about speed and more about control, from camera movement and lighting to characters, motion, and lip sync. And they explore what those capabilities could mean for professional creative workflows, where artists and studios need predictable tools rather than simply generating a video from a prompt.
Resources:
Follow Gorkem Yurtseven on X: https://x.com/gorkem
Follow Batuhan Taskaya on X: https://x.com/isidentical
Learn more about fal: https://fal.ai
Follow Jennifer Li on X: https://x.com/JenniferHli
Stay Updated:
Find a16z on YouTube: YouTube
Find a16z on X
Find a16z on LinkedIn
Listen to the a16z Show on Spotify
Listen to the a16z Show on Apple Podcasts
Follow our host: https://twitter.com/eriktorenberg
Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.
Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.