Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Introducing fal's H3 Max Director, the first natively continuous real-time frontier video model for action-controlled long-form generation. Unlike H3 Max and other video models, H3 Max Director generates a single continuous video stream, maintaining context over time rather than chaining together discrete clips. It’s the technology behind our experimental... show more

fal

65,392 subscribers

516,450 views • 8 days ago •via X (Twitter)

Science & Technology Education

Anya Rossi• Live Now

Private livecam show

35 Comments

fal8 days ago

Use H3 Max Director to build continuous livestreams, interactive video, persistent worlds, or applications we haven’t thought of yet. Try it now.

fal8 days ago

See H3 Max Director in action on where it’s powering multiple channels of continuous, interactive AI video in real time. Tune in:

Toomas B8 days ago

asked fable 5.1 to use h3 max director to create a the office sketch, oneshotted this

Levan Kvirkvelia8 days ago

EVERY DAY THERE IS A NEW RELEASE WTF FAL RESEARCHERS

Gorkem Yurtseven8 days ago

I can guarantee this is the most fun you'll have on the internet today

cramer8 days ago

Somebody cooked here

rv4life8 days ago

Yeah, absolute nightmare fuel. And not the prompt.

noah8 days ago

It's a good model sir

Tina Sang8 days ago

fal shipping faster than the speed of light

joe atkins8 days ago

fal team

Blendi8 days ago

how i be with h3 max director

Dy Co8 days ago

So they're basically just using what the open source community made 3 weeks ago, repurposing it for their model and letting people pay for it. Nice. At this point a different group will release a better faster open source alternative to H3 Max and it'll be better.

Create a Meme8 days ago

Dang! Fal you guys are absolutely cooking with all these AI breakthroughs. 🤯 👨‍🍳👩‍🍳 @AIMemeB0t can you make a meme celebrating all these new models and features? 🔥🤖

maharshi8 days ago

LFGGGG!!!

RAZA | AI EXPLORER8 days ago

Continuous real-time video generation is a huge step forward. Excited to see what creators build with H3 Max Director.

kiri498 days ago

who cares about the API, make your M3 Max models open weights like the model you based it on

Elsi Hametaj8 days ago

"...the first natively continuous real-time frontier video model for action-controlled long-form generation." You guys really wanted a worlds first record, didn't you.

Ege Beşe8 days ago

wow, so many possibilities now

GifCo7 days ago

This looks like absolute dog shit after about 30-40s. How can you say this is continuous!! More click bait bullshit from Fal

Artificial Guy / João Vitor A.8 days ago

impressive lol

chefboirdoge8 days ago

Ok now open ai buy them and merge with sora and give in sub lol

Artur Kre8 days ago

me watching this 🤤 while my fal credits melt away 💸💸💸

poz8 days ago

open the weights

Edets Com8 days ago

Damn— Got me up at midnight trying to reverse engineer the prompt for this masterpiece.

AHU7 days ago

Does it support some reference images?

Kevin8 days ago

I bought credits but it keeps saying "User is locked. Reason: Exhausted balance. Top up your balance" even though I have enough credits

Glitter Gal8 days ago

Darn okay, impressive.

adrián8 days ago

can't wait to try it and hopefully implement it into my app (SlopTime 👀)

Pravesh Jha8 days ago

you guys are cooking!

Finance Cyprus7 days ago

Cost for 1 minute video with discount?

𝕻 𝕵 𝖆 𝖄®8 days ago

But you can’t use your own images?

Bk1man8 days ago

Worth adding: the H3 base fal fine-tuned sits at 5.09M downloads on HF, while the MiniMaxAI org still lists zero H3-Max weight repos — just the 08-13 base. Director is a strong product, but the open-weights ask filling your own replies is the real story.

The Bentist (Dr. Winters)8 days ago

NGL you guys kinda been cookin lately

kriGeta7 days ago

Any possibility for open weights? Please dont ruin them before releasing.

Goran k6 days ago

Bro imagine being the only one with access to this in 2015

Related Videos

H3 Max Director 1.1 is now available. New features for longer, more controllable video generation: - Start with any frame and insert keyframes throughout a generation - Introduce audio at any point, with native audio conditioning - Generate natively in 1080p - Run continuous generations for up to 15 minutes Director adapts to new visual and audio inputs in real time, giving you precise control without breaking continuity.

H3 Max Director 1.1 is now available. New features for longer, more controllable video generation: - Start with any frame and insert keyframes throughout a generation - Introduce audio at any point, with native audio conditioning - Generate natively in 1080p - Run continuous generations for up to 15 minutes Director adapts to new visual and audio inputs in real time, giving you precise control without breaking continuity.

fal

73,896 views • 3 days ago

One image. Every angle. Rendered under 3 seconds. Multi-angle is now live for H3 Max on fal Set the horizontal and vertical camera angle in degrees and H3 Max moves through the scene from a single viewpoint. Shape, position, and materials stay put as the perspective changes, so the angles read as the same moment rather than separate generations.

One image. Every angle. Rendered under 3 seconds. Multi-angle is now live for H3 Max on fal Set the horizontal and vertical camera angle in degrees and H3 Max moves through the scene from a single viewpoint. Shape, position, and materials stay put as the perspective changes, so the angles read as the same moment rather than separate generations.

fal

89,005 views • 2 days ago

i made an infinite ai shopping channel where you invent the products and our host graham has to sell them built with H3 Max turbo, realtime 1080p for the first time

i made an infinite ai shopping channel where you invent the products and our host graham has to sell them built with H3 Max turbo, realtime 1080p for the first time

Blendi

37,264 views • 4 days ago

PixVerse just launched their first real-time video world model R1, powered by fal's high-performance compute. It turns AI video from static clips into a real-time, interactive world. 🔄 Continuous generation stream 🧠 Autoregressive memory system ⚡ Ultra-fast response engine

PixVerse just launched their first real-time video world model R1, powered by fal's high-performance compute. It turns AI video from static clips into a real-time, interactive world. 🔄 Continuous generation stream 🧠 Autoregressive memory system ⚡ Ultra-fast response engine

fal

19,243 views • 7 months ago

Meet Wildlife World From rainforest canopies to coral reefs, you vote on where we go next AI-generated wildlife 24/7 powered by H3 Max Director. Watch live:

Meet Wildlife World From rainforest canopies to coral reefs, you vote on where we go next AI-generated wildlife 24/7 powered by H3 Max Director. Watch live:

fal

14,165 views • 4 days ago

built this game with fable 5.1 + fal H3 max director > you enter a city of fruits (fruitville) > you decide what happens next > every choice changes the characters, relationships and story > new characters, events and locations can appear at any time > the world remembers what happened and keeps evolving > every scene is generated in real time the game is generated as you play it

built this game with fable 5.1 + fal H3 max director > you enter a city of fruits (fruitville) > you decide what happens next > every choice changes the characters, relationships and story > new characters, events and locations can appear at any time > the world remembers what happened and keeps evolving > every scene is generated in real time the game is generated as you play it

Matthieu

29,912 views • 8 days ago

is back! It now runs continuously, with LLM-generated prompts that anyone watching can upvote to decide what gets created next. We’ve been experimenting with what faster-than-real-time AI models like H3 Max make possible for entirely new forms of content and entertainment. is one experiment in what that future could look like. We’ll keep evolving it as more people use it by adding new channels, improving the experience, and seeing where people take it. You can also submit ideas for new channels or apply to stream your own.

is back! It now runs continuously, with LLM-generated prompts that anyone watching can upvote to decide what gets created next. We’ve been experimenting with what faster-than-real-time AI models like H3 Max make possible for entirely new forms of content and entertainment. is one experiment in what that future could look like. We’ll keep evolving it as more people use it by adding new channels, improving the experience, and seeing where people take it. You can also submit ideas for new channels or apply to stream your own.

fal

75,387 views • 11 days ago

ok promising (but rough) start on an interactive video project using fal's H3 Max (and now Max Turbo) basically it's a first person POV 'date simulator' but the woman across from you is a secret agent using Blendi's basic interactive frame work (and a lot of other additional fable 5.1 work) each run is unique, depending on your choices throughout the dinner lots of issues with physics and drift but... there's def some sort of new medium here. and it's kind of magic to see the video just keep going in real time. trying to get calls as tight as possible (calls claude for story, nanobanana for facial consistency, h3 max for three different clips so they're ready to go when you select your path) and you can see my timing and when things trigger in the bottom right of the screen shout out to Tina Sang and the fal team for dropping me a few extra credits to keep working on this

ok promising (but rough) start on an interactive video project using fal's H3 Max (and now Max Turbo) basically it's a first person POV 'date simulator' but the woman across from you is a secret agent using Blendi's basic interactive frame work (and a lot of other additional fable 5.1 work) each run is unique, depending on your choices throughout the dinner lots of issues with physics and drift but... there's def some sort of new medium here. and it's kind of magic to see the video just keep going in real time. trying to get calls as tight as possible (calls claude for story, nanobanana for facial consistency, h3 max for three different clips so they're ready to go when you select your path) and you can see my timing and when things trigger in the bottom right of the screen shout out to Tina Sang and the fal team for dropping me a few extra credits to keep working on this

Gavin Purcell

23,970 views • 9 days ago

Okay I built it! 🍰 Infinite Slop An infinite and interactive AI generated live stream of slop that goes on forever and ever Anything that you write in the chat is generated next and AI will try to connect it to the previous video so there's an actual sequence and story line Idea by marc-antoine and Rehan Sheikh's original stream so I made it It'd be very expensive to run this so fal very generously sponsors this They also fine-tuned the model that makes this possible for the first time, making Minimax H3 50x faster and as "Max" and it can now generate videos faster than you can watch them, very cool! Let me know what you think!!!! 😊

Okay I built it! 🍰 Infinite Slop An infinite and interactive AI generated live stream of slop that goes on forever and ever Anything that you write in the chat is generated next and AI will try to connect it to the previous video so there's an actual sequence and story line Idea by marc-antoine and Rehan Sheikh's original stream so I made it It'd be very expensive to run this so fal very generously sponsors this They also fine-tuned the model that makes this possible for the first time, making Minimax H3 50x faster and as "Max" and it can now generate videos faster than you can watch them, very cool! Let me know what you think!!!! 😊

@levelsio

2,590,921 views • 13 days ago

A breakthrough in real-time video generation. As a research preview developed with NVIDIA and shared at NVIDIAGTC this week, we trained a new real-time video model running on Vera Rubin. HD videos generate instantly, with time-to-first-frame under 100ms. Unlocking an entirely new creative paradigm and bolstering the foundations of our General World Model, GWM-1. Real-time generation opens a fundamentally different design space for video models and world simulation. We're investing in co-designing our models alongside advances in hardware to keep pushing this frontier.

A breakthrough in real-time video generation. As a research preview developed with NVIDIA and shared at NVIDIAGTC this week, we trained a new real-time video model running on Vera Rubin. HD videos generate instantly, with time-to-first-frame under 100ms. Unlocking an entirely new creative paradigm and bolstering the foundations of our General World Model, GWM-1. Real-time generation opens a fundamentally different design space for video models and world simulation. We're investing in co-designing our models alongside advances in hardware to keep pushing this frontier.

Runway

1,164,245 views • 5 months ago

Introducing The Matrix --- a foundation world model for generating infinite-length, hyper-realistic videos with real-time, frame-level control: - Infinite-length video generation - 720p high-quality rendering - Real-time, frame-level control at 16 FPS - Generalization to real-world video control 🔗Blog: 📄Paper: 💻Code & Playable Demo: Coming soon! Key Innovation: A brand new technique called the shift-window denoise process model, enabling auto-regressive generation for diffusion and consistency models in real-time. Special thanks to project leader Ruili Feng and the entire Matrix team for their dedication and hard work over the year-long project.

Hongyang Zhang

178,407 views • 1 year ago

I did it, the first-AI-generated collective video game. The chat choose what happenned next and then they vote. Still so much room for improvement, I need to fix the context situation, the dialogue, the creativty, and being able to better manipulate the state but the possibility are awesome !!! I can’t even imagine what will be possible in just a few months. Can’t wait. Huge thanks to Tina Sang and fal for giving me the opportunity to push the limits of this awesome H3 Max technology. 🚀

I did it, the first-AI-generated collective video game. The chat choose what happenned next and then they vote. Still so much room for improvement, I need to fix the context situation, the dialogue, the creativty, and being able to better manipulate the state but the possibility are awesome !!! I can’t even imagine what will be possible in just a few months. Can’t wait. Huge thanks to Tina Sang and fal for giving me the opportunity to push the limits of this awesome H3 Max technology. 🚀

Defend Intelligence (Anis Ayari)

21,819 views • 9 days ago

SITUATION EXPLAINED: MiniMax released H3 with open weights, and it's already #1 in video editing. • Priced at $7.80 per minute of 2K video, against Seedance 2.0 at $22.45 and Kling 3.0 at $20.16 • It ranks #1 on Artificial Analysis for video editing, #2 for text-to-video behind Gemini Omni Flash, and #3 for image-to-video • H3 takes text, images, video, and audio in a single context and generates 5-15 second clips at native 2K, 24fps, with native stereo audio • MiniMax is releasing the weights under its Community License within days, which would make it the strongest open-weights video model by a wide margin • It launched the same week ByteDance rolled out Seedance 2.5, which now generates 30-second 4K video from a single prompt • Seedance 2.5's face consistency across a full 30 seconds is the visible jump, earlier models drifted into a different face partway through • Voices are still the obvious tell, the video reads as real long before the audio does • Smaller labs building purpose-built video models are now beating frontier labs at this specific task Theo Jaffee: "This is gonna be one of the biggest examples of things that people are gonna do with these models, historical simulations, basically. Nostalgia videos for zoomers."

SITUATION EXPLAINED: MiniMax released H3 with open weights, and it's already #1 in video editing. • Priced at $7.80 per minute of 2K video, against Seedance 2.0 at $22.45 and Kling 3.0 at $20.16 • It ranks #1 on Artificial Analysis for video editing, #2 for text-to-video behind Gemini Omni Flash, and #3 for image-to-video • H3 takes text, images, video, and audio in a single context and generates 5-15 second clips at native 2K, 24fps, with native stereo audio • MiniMax is releasing the weights under its Community License within days, which would make it the strongest open-weights video model by a wide margin • It launched the same week ByteDance rolled out Seedance 2.5, which now generates 30-second 4K video from a single prompt • Seedance 2.5's face consistency across a full 30 seconds is the visible jump, earlier models drifted into a different face partway through • Voices are still the obvious tell, the video reads as real long before the audio does • Smaller labs building purpose-built video models are now beating frontier labs at this specific task Theo Jaffee: "This is gonna be one of the biggest examples of things that people are gonna do with these models, historical simulations, basically. Nostalgia videos for zoomers."

MTS

55,187 views • 1 month ago

I might’ve found a way to make yesterday’s Blender to Real video ~5,000x faster (and better imo) lol. I put the scene together with GPT-6 Astra and PATINA in about 30 minutes. But it had around 7 million polygons, so rendering took a while. I split it into chunks across 8 GPUs and rendered it in about 30min. My estimate for an M4 Pro was around 40 hours. Obviously, I could’ve optimized the settings, but that’s for the version I posted yesterday. Since then, I’ve been trying different prompts to get H3 Max to follow the Blender animation while making the buildings and materials actually look real. I think I finally got it. The 30s video took about 30s to generate with H3 Max. Compared to that estimated 41h render, that’s roughly 5000x faster :) The part I’m most excited about is how much control this gives me. GPT-6 Astra can build a detailed 3D prototype in minutes, and I can use Blender to set up the camera and movement however I want, then have H3 Max follow it.

I might’ve found a way to make yesterday’s Blender to Real video ~5,000x faster (and better imo) lol. I put the scene together with GPT-6 Astra and PATINA in about 30 minutes. But it had around 7 million polygons, so rendering took a while. I split it into chunks across 8 GPUs and rendered it in about 30min. My estimate for an M4 Pro was around 40 hours. Obviously, I could’ve optimized the settings, but that’s for the version I posted yesterday. Since then, I’ve been trying different prompts to get H3 Max to follow the Blender animation while making the buildings and materials actually look real. I think I finally got it. The 30s video took about 30s to generate with H3 Max. Compared to that estimated 41h render, that’s roughly 5000x faster :) The part I’m most excited about is how much control this gives me. GPT-6 Astra can build a detailed 3D prototype in minutes, and I can use Blender to set up the camera and movement however I want, then have H3 Max follow it.

ilker

56,896 views • 5 days ago

Gemini Robotics ER 2 is our most capable embodied reasoning model designed for physical AI 🤖 Built as a high-level brain for robotics, the model connects directly to the Gemini Live API. It processes continuous video streams to track progress, call tools, search the web, and command an action model in real time. Upgrades for developers include: + Enhanced progress tracking + Real-time mid-execution failure detection + Multi-robot collaboration + Advanced safety instruction following

Gemini Robotics ER 2 is our most capable embodied reasoning model designed for physical AI 🤖 Built as a high-level brain for robotics, the model connects directly to the Gemini Live API. It processes continuous video streams to track progress, call tools, search the web, and command an action model in real time. Upgrades for developers include: + Enhanced progress tracking + Real-time mid-execution failure detection + Multi-robot collaboration + Advanced safety instruction following

Google AI Developers

21,498 views • 1 month ago

Real-time, playable worlds from a single sentence. Alibaba Cloud Happy Oyster 1.0 is now available on fal, streaming in our new Realtime Playground. - Enter, explore, and interact with a world as it generates - Adventure: first or third person, real-time movement and camera control - Directing: inject commands mid-stream, with 3-minute narrative consistency

Real-time, playable worlds from a single sentence. Alibaba Cloud Happy Oyster 1.0 is now available on fal, streaming in our new Realtime Playground. - Enter, explore, and interact with a world as it generates - Adventure: first or third person, real-time movement and camera control - Directing: inject commands mid-stream, with 3-minute narrative consistency

fal

88,208 views • 1 month ago

Today, we're announcing the next evolution of Sword Thrive: a fully revamped experience anchored by a new conversational care model, Vision AI, and a more powerful Phoenix. At the center of this evolution is a first-of-its-kind care model: a 3-way chat connecting the member, Phoenix, and their Doctor of Physical Therapy in a single interface, making continuous, 24/7 care possible for the first time, with the clinical oversight only a human specialist can provide. Vision AI, our proprietary technology trained on years of movement data, now delivers precise movement detection in real-world home environments - cluttered, noisy, and everything in between - and runs entirely through the Sword Health app on smartphones and tablets. Phoenix, our AI Care Specialist, is now more capable as well, providing deeper, personalized two-way conversational guidance and real-time form correction, adapting to each member as they move. From proactive care to continuous recovery, this is how healthcare should show up: with reliability, clinical memory, and real-time support that never drops the thread. Watch the video to see it in action:

Today, we're announcing the next evolution of Sword Thrive: a fully revamped experience anchored by a new conversational care model, Vision AI, and a more powerful Phoenix. At the center of this evolution is a first-of-its-kind care model: a 3-way chat connecting the member, Phoenix, and their Doctor of Physical Therapy in a single interface, making continuous, 24/7 care possible for the first time, with the clinical oversight only a human specialist can provide. Vision AI, our proprietary technology trained on years of movement data, now delivers precise movement detection in real-world home environments - cluttered, noisy, and everything in between - and runs entirely through the Sword Health app on smartphones and tablets. Phoenix, our AI Care Specialist, is now more capable as well, providing deeper, personalized two-way conversational guidance and real-time form correction, adapting to each member as they move. From proactive care to continuous recovery, this is how healthcare should show up: with reliability, clinical memory, and real-time support that never drops the thread. Watch the video to see it in action:

Sword Health

27,974 views • 6 months ago

We're changing our name. Captions is now Mirage. Today's most popular medium is short-form video. It's the fastest-growing format, it’s what people want, and it’s what brands need to grow. But unlike other forms of AI content like text or image, which can be generated and ready-to-use with a single prompt, generated video still requires a lot of manual work and editing to be ready to share. This is where the real opportunity lies. The way we see it, the real race for AI video hasn’t even begun. The world is waiting for a foundation model for video that can take an idea and produce a fully usable, cohesive output from scratch. So that’s what we’re building. We’re rebranding to Mirage to reflect our expanded vision for video, starting with the short-form format, through frontier AI research and models. Check it out at mirage dot app

We're changing our name. Captions is now Mirage. Today's most popular medium is short-form video. It's the fastest-growing format, it’s what people want, and it’s what brands need to grow. But unlike other forms of AI content like text or image, which can be generated and ready-to-use with a single prompt, generated video still requires a lot of manual work and editing to be ready to share. This is where the real opportunity lies. The way we see it, the real race for AI video hasn’t even begun. The world is waiting for a foundation model for video that can take an idea and produce a fully usable, cohesive output from scratch. So that’s what we’re building. We’re rebranding to Mirage to reflect our expanded vision for video, starting with the short-form format, through frontier AI research and models. Check it out at mirage dot app

Gaurav Misra

137,555 views • 1 year ago

world models are here! fal is the best platform for real-time audio and video inference. RT for access 😇 incredible work by Eloi Alonso to open source this model

world models are here! fal is the best platform for real-time audio and video inference. RT for access 😇 incredible work by Eloi Alonso to open source this model

fal

123,318 views • 1 year ago