We're bringing online 1 GW of net new compute in the next 24 months, and I'm hiring for Tactical Compute Ops in SF to help make it happen.
You'll work directly with me and @nina_gerov, one of the best ops leaders I know.
Finance/IB, Legal, and other backgrounds are more than welcome.
DM me with a few lines on yourself and why you'd want to join.
Show more
We've been pushing diffusion RL into new verifiable tasks. The surprise: lip-sync is one of them.
The result is the highest-quality, fastest, and cheapest lip-sync model on the planet. Built on top of H3 Max by @fal
Show more
Who in generative media (image/video/audio/3D/robotics) fields needs a shit ton of hoppers right now?
not in a few months/weeks etc. literally now. Hit me up in DMs.
@aollivier82 @levelsio it actually costs $864 with the new price :)
$0.01 * 60 * 60 * 24 = $864.
less than $1K
Realtime Interactive Education with H3 Max
this model is revolutionary
Fastest reference-to-video model with frontier quality. And the nice thing is, it will only get faster from here. Pair your character with a @suno generated music and get super fun music videos under a 30 seconds!
Show more
Reference-to-Video is now live for MiniMax H3 Max on fal!
This is an early preview that can achieve up to RTF=1 (and with references around RTF=1.5) on 768p, and we aim to improve the speed by 2x while slashing costs during this week. Quality wise it delivers the same performance you come to expect from H3 Max.
Show more
Try out now yourself thanks to some help from the folks @fal
Free till I run out of credits
I built the production version fully from my phone using codex remote to my laptop in SF while I’m in NYC
Still working out some kinks, leave feedback
Show more
if you're a creator and you want to build something with h3 max live, comment below and we'll try and support!
❤️❤️ MiniMax's official consumer app is now powered by @fal
Experience the speed on MiniMax Design
H3 Max is now available on MiniMax Design!
#MiniMaxH3# #MiniMaxDesign# #MiniMax#
⚡ Generate faster than ever
💰 480p starting at just $0.02/sec
🆓 Get 3 free generations through September 3
From one prompt to the next scene, keep creating at the speed of imagination.
Show more
H3 began as a video model. Now it’s learning to carry new stories forward. 🏃
we built live TV controlled by the audience
pick a channel → suggest what happens next → watch it become the next scene
We didnt sleep to make that happen !! Its clearly incredible to have the luck to work on something so innovative ! have a look, choose your history, make whatever you want !
PS: This is powered by H3 Max Director, an auto-regressive and natively continous version of H3 Max.
The generated videos tie in better than what T2V/I2V can do, with a context window of up to 2 minutes. NATIVELY.
This is very experimental, we will be improving quality a lot during this week but also showing what is possible with H3 Max.
Show more
Built an infinite scrolling tiktok app with @fal 's H3 max model
Only took a few hours with @OpenAI Codex including the demo screen recording fully autonomously
It generates content on the fly as you scroll, and even can extend videos as you watch them up to 30s
Pretty wild
Show more
“The core technology is MiniMax H3's accelerated version H3 Max, with post-training + inference optimization by @fal:
1. About 5-second video generated in under 3 seconds
2. 15-second video output in about 9 seconds
3. Throughput an order of magnitude higher than official H3
This means AIGC is no longer just produce-first-then-post; it can now enter live streams in real time”
Show more
This is a live demo from the future!!! 🤯🤯🤯
An interactive sitcom? A playable episode?
Games and movies are merging into one thing. I genuinely cannot stop playing this.
Real-time video generation will change the entertainment industry forever. We are so close.
Thanks to @fal team . @Hailuo_AI
Show more
@JulienBlanchon @fal Hi! We are still working on it! Probably couple weeks away, we want to make sure it is optimized in a variety of hardware types before we release it
Token market fit!
Generative media and coding.
We released multiple times the SoTA image model in open source. With similar speed benefits. All you need to do is check our blog and huggingface.
We even go as deep into kernel work and training speculative decoding adapters for prompt expansion.
Be sure H3 Max will come
Show more
Synthetic video and image generation represents the largest rapidly emerging TAM for token consumption after coding.
