Rehan Sheikh connected MiniMax H3 Max to a Twitch livestream, showing a channel supplied by newly generated video.

Pieter Levels built Infinite Slop, where the site's chat helps decide what airs next.

The speed makes these experiments possible. fal says H3 Max, its post-trained version of MiniMax H3, can generate five seconds of video in under three seconds.

Those are fal's reported results. But the timing explains the idea: if generation can stay ahead of playback, the audience does not have to wait between every clip.

Now connect that speed to a story, viewer choices or a recommendation algorithm.

I am Alex. Welcome to ShortCu8 by Innov8.

Lets Dive Deep🐰🐰.

Separate keeping the stream running from deciding what happens next.

Fast generation helps with the first. A prompt, a vote or a prediction about the viewer can handle the second.

That distinction helps us read these demos. An endless channel and a personalised movie need different systems, even if they use the same video model.

Austin Hurwitz's article explores what happens when media keeps being produced while people experience it.

Henry Daubrez has already shown an early version of this idea.

His THIS WAY demo uses H3 Max to generate two possible continuations ahead of an audience vote. The chosen branch becomes part of the ongoing story. He says the system tracks characters, story arcs and consequences.

In his replies, Henry explains that the storyline has planned beats and an intended endpoint. Audience choices change the route, while the plot is meant to converge.

This is a working demo by his description, not proof of an unlimited interactive movie.

Still, imagine watching a thriller with friends and voting on whether the protagonist opens the door or follows the noise upstairs. Upcoming scenes could be generated during the session, rather than filming every branch before release.

Watching together would include arguing about what should happen, then seeing the consequences.

Imagine a fictional Kochi apartment sitcom with you and your friends as recurring characters, using everyone's likeness with permission.

You could suggest a situation or speak to a character. A memory system could carry earlier events into the next episode, including the running joke about who never pays their share of the rent.

Or keep yourself out of the cast and let the show adapt to your taste. Your version could spend more time with one character, while a friend's version follows another.

These are possibilities, not capabilities established by the livestream demos. They would need reliable character continuity and memory across sessions.

But they show what faster generation could support: entertainment whose next episode does not need to be completely finished before you arrive.

The more unsettling possibility would not require you to vote or type a prompt.

TikTok already explains how viewing behaviour and other interactions help its system predict which existing videos someone might find interesting.

Now imagine connecting those predictions to a generator.

Watch -> observe the response -> predict the next segment -> generate -> repeat.

The system could try different premises or pacing, then use the response to inform later generations. It would be estimating interest, not reading your mind.

Today, a recommendation system has to find suitable material in its available catalogue. A generative feed could attempt to make something matching that prediction.

The internet already contains more video than we can finish. The additional possibility is a supply that keeps adjusting to one viewer.

That could become a personal entertainment channel. With watch time as its main objective, it could also become an attention machine that keeps testing ways to hold us there.

The goal given to a generative feed would matter as much as its ability to produce another clip.

Generated video can already supply a stream, and builders are experimenting with chat input and voting over generated story branches.

That is narrower than a movie reacting instantly to anything you do. Preparing clips ahead of time introduces a delay between a new input and its visible effect. Generating alternative branches also spends compute on scenes viewers may never see.

Full-length coherent stories and a profitable personalised feed remain unproven here.

Fast rendering supplies the clips. The system around it decides which ones we see.

Now go and test H3Max on Fal

We read every reply - just reply to this email and let us know how we can improve !

#AD1

Usage-based and hybrid pricing models are changing how B2B companies generate revenue — and creating new headaches for the finance teams behind them.

Tabs co-founder Rebecca Schwartz and PwC Partner Amit Dhir sat down to unpack exactly what that means in practice: how pricing model decisions ripple into revenue recognition, forecasting, and financial ops — and what it takes to scale without piling on manual work.

Watch the on-demand recording to get practical frameworks, real-world examples, and a clear path to operationalizing usage-based revenue — including a forward-looking take on how AI will reshape financial workflows. If your team is navigating pricing complexity heading into the back half of the year, this is worth an hour.

Watch the recording now

#AD2

The difference between growing stores and stagnant ones isn't more effort. It's better insights. StoreClaw analyzes your Shopify and Amazon data, surfaces your biggest growth opportunities, and helps you increase GMV while protecting profit. Start free with bonus tokens. No credit card required.

Try it FREE today