I was a strong advocate. The platform and tech are genuinely good. That makes this more disappointing.
A year ago I activated $2,000 in credits offered by the Fal team at a hackathon they were sponsoring. Over that year I spent somewhere between $200 and $300. Last week, with no email, no warning, no communication of any kind, the remaining balance was wiped.
Credit expiry isn't unusual. Not communicating it is. Only when I reached out myself did they confirm what had happened, tell me nothing could be done, and point to a small notice buried somewhere in the platform. That's it.
No proactive comms. No goodwill. No attempt to retain a developer who'd built multiple apps on their platform and intended to build more. In a space where developer trust is everything, that reads less like an oversight and more like a policy.
My advice: the tech is worth exploring. The support is not. If developer community matters to you, look for a platform where it matters to them too.
H3 Max is fal's post-trained take on MiniMax H3, same open-weight base model, but retrained for better prompt adherence and visual quality, then tuned hard for speed.
The problem it's solving: Video gen usually makes you pick one. Want quality? Wait longer. Want speed? Accept worse output. H3 Max tries to break that tradeoff instead of accepting it.
How: fal Research fed it new training data focused specifically on prompt understanding and aesthetics, while their inference team rebuilt the serving stack around it (running on NVIDIA GB200 NVL72) so the speed gains don't eat into quality.
Does it work? They ran it against 12 other video models, including the original H3, Gemini Omni Flash, Wan 3.0, Seedance 2.5, Kling 3, Veo 3.1, using human preference scoring (Bayesian Elo). H3 Max came out #1 on overall quality, prompt understanding, and aesthetics, and won most head-to-head matchups it was put in. Independent benchmarks (Artificial Analysis, Design Arena) back this up too.
Key features:
Text-to-Video and Image-to-Video endpoints
5-second clip in about 3 seconds, that's 35x the throughput of the official H3 endpoint
Usable via Playground, fal Agent, or API
Good fit if you're building anything that needs fast, high-quality video generation at scale, not just a one-off demo. Try it: Text to Video · Image to Video
P.S. I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified → @rohanrecommends