WATCH AI STREAMSH3 MAX VIDEO SITES DIRECTORY
Discover public video sites built with H3 Max—from continuous AI TV and audience-directed stories to interactive games and on-demand shows. Browse the directory, compare how each site works and visit the original experience.
FEATURED LIVE FORMATS / 003
Three live formats lead the directory. Broadcast activity remains separate from website availability.
View profile Everyone watches the same stream.
Choose Infinite Slop when you want shared chaos. Everyone sees the same public feed, while chat submissions and audience voting shape the queue. Your idea competes with other viewers before it reaches the screen, so the experience feels closer to community-operated television than a private generation session.
View profile Multiple worlds, one audience-directed network.
Choose fal.live when you want recurring worlds instead of one chaotic feed. Its themed channels give the audience a shared setting, then use proposals and voting to decide which direction the next generated scene should take. The structure makes it easier to follow a format while still changing the story together.
View profile Create a contestant. Enter the story.
Choose Renoise Live when identity and character direction matter more than a public prompt queue. Participants can create a contestant, enter a shared survival-show setting and direct character actions. It is the closest of the three experiments to a game, although the resulting story still plays as generated television.
BROWSE / 09 SITES
Browse public websites that use H3 Max to generate, stream or power interactive video. Filter by format, then open a profile to see how the experience works and where its model evidence comes from.
View profile A shared AI channel where public prompts and voting decide what reaches the screen next.
View profile A network of recurring AI television worlds shaped by audience proposals and votes.
View profile An AI survival show where participants create contestants and direct character actions.
View profile One public canvas where people propose, vote on and steer each generated scene.
View profile An infinite-show platform where watching is free and paid steering can fund upcoming clips.
View profile A compact prototype that turns one viewer choice into the next 5–15 second video scene.
View profile A persistent point-and-click adventure where each choice grows a saved video branch.
View profile A multiplayer drawing game where H3 Max transforms sketches while other players guess.
View profile A streaming-style catalog that turns a premise into an AI-generated movie or series.
CHOOSE YOUR CONTROL
The biggest difference between these experiments is not image quality. It is how much control the viewer receives, whether the world persists and whether participation affects a shared queue or a specific character.
Choose a shared broadcast if you enjoy a public queue where every viewer sees the same result. Prompts and votes make the audience part of the programming process, but any one suggestion can lose to another.
Choose a recurring world if you want the next scene to change without discarding the whole format. A stable theme gives generated clips more context and makes it easier to follow what the room is directing.
Choose a directed character if you want your identity or actions represented inside the show. This creates more personal investment, but it also requires the product to manage a cast, relationships and continuity across scenes.
THE LOOP CLOSES
When the next clip can be generated before the current clip finishes, video becomes part of a continuous feedback loop.
fal reports that the model can generate a five-second 768p video in under three seconds. This is fal's published performance claim, not an independent Watch AI Streams benchmark. It describes model inference time rather than the full delay from a viewer pressing send to a new scene reaching the player.
Real broadcasts still have to process prompts, run safety checks, wait for available compute and deliver the result. A short generation time creates room for those steps; it does not guarantee a gap-free stream. That is why this directory treats model capability, website access and active broadcast status as separate facts.
fal performance source · Aug 27, 2026fal official performance claim · not independently benchmarked
H3 MAX LIVE STREAM COST
fal bills standard video output by the generated second. This calculator multiplies the selected resolution rate by broadcast hours and channel count. It estimates raw text-to-video generation only; prompt processing, reference inputs, safety checks, retries, moderation, storage and delivery are not included.
Pricing source · verified Sep 8, 2026VERIFICATION, NOT VIBES
Every listed product needs a public video experience, attributable H3 Max evidence and a dated review. Missing evidence stays visible instead of becoming a confident badge.
Read the full methodologyPeople must be able to watch, create or interact with video on a public website. Source repositories, API documentation without a viewing product and social accounts that only post finished clips are outside this directory.
A site is labeled with H3 Max only when its operator, interface, public API, browser client or creator identifies the model. Similar visual style is not enough. The evidence level, source and review date stay attached to the record.
A successful page load only proves that the website is reachable. LIVE requires an observable broadcast signal, and non-live products do not receive a broadcast field at all. Missing details remain Unknown or Unconfirmed.
H3 MAX DIRECTORY FAQ
No. It is a video generation model post-trained by fal from the open-weight MiniMax H3 base. Separate products add viewing, interaction, moderation, queues, world state and delivery. This directory covers public websites built around those different product layers.
The featured live section covers Infinite Slop, fal.live and Renoise Live. The full directory also includes DirectedByPublic, Hero TV and non-live H3 Max experiences. Each profile explains the available controls before linking to the original website.
fal reports that the model can generate a five-second 768p video in under three seconds. That is a company-published performance claim, not an independent benchmark by Watch AI Streams. Prompt expansion, safety checks, queues, storage and delivery can add time before a viewer sees the next scene.
fal currently lists standard output pricing of $0.05 per second at 480p and $0.08 per second at 768p. A continuous channel also needs orchestration, retries, storage, bandwidth and moderation. The calculator estimates raw text-to-video output only and displays the date when pricing was last checked.
It depends on the site. Some use prompts and voting, some let visitors steer characters or choose branches, and Renderdown turns drawing into a multiplayer clue. Slopflix is different again: users commission an on-demand title instead of directing every scene.
Not necessarily. A website can remain reachable while its broadcast is paused or impossible to verify. We therefore report website availability and broadcast activity separately. If there is no reliable public signal that new video is being generated or played, the broadcast status remains UNKNOWN.
OPEN CALL