A single endless AI channel where the chat queue decides the next fifteen seconds.
Audience-Directed ShowH3 MaxPrompt QueueAudience Voting
Verified
Model brief
H3 Max
H3 Max is fal's post-trained version of MiniMax H3, tuned for prompt adherence and visual quality and served on an inference stack co-designed with the model. fal states that it produces a five-second clip in roughly three seconds, which is the property that makes continuous, audience-driven channels practical.
Facts
- Developer
- fal
- Type
- Video generation (post-trained variant)
- Released
- 27 Aug 2026
- Native audio
- Not publicly disclosed
- Real-time capable
- Yes
- Open weights
- Not publicly disclosed
- Public API
- Yes
- Based on
- MiniMax H3
- Tracked deployments
- 6
- Last verified
It is available through fal's playground and API. Whether the post-trained weights are published is not stated, so its open-source status is recorded as unknown.
Deployments
Platforms running on H3 Max
Role: Primary generation
Always-on AI television where the audience votes on the next scene.
Audience-Directed ShowH3 MaxPrompt QueueAudience Voting
Verified
Role: Primary generation
Directed By Public
Viewers propose and vote on scene seeds; the winning seed drives a claimed real-time canvas — not red-LIVE until a public detector exists.
Audience-Directed ShowH3 MaxFree-form PromptAudience Voting
Verified
Role: Primary generation
H3 Story Lab
Make one choice and watch H3 MAX generate the next story scene — on demand, not a continuous LIVE feed.
Interactive VideoH3 MaxMultiple Choice
Verified
Role: Primary generation
Renderdown
A browser generative playable experiment on Vercel — on-demand interaction, not a broadcast LIVE badge.
Interactive WorldH3 MaxMulti-player Shared Input
Verified
Role: Primary generation
A starting image becomes a world you steer by typing or speaking — on demand, not a live broadcast.
Interactive WorldH3 MaxFree-form PromptVoice Input
Verified
Role: Primary generation
Recent updates
Timeline
- · New Launch
Wildlife World, Artificial TV, Directed By Public, H3 Story Lab, Renderdown, ALIBI and APOB Live were added as published directory pages with conservative LIVE rules.
- · New Launch
Pieter Levels published Infinite Slop, a continuous AI video channel that generates the next clip from the top-voted chat prompt. A dedicated domain followed the next day.
- · Model Change
fal announced H3 Max, a post-trained variant of MiniMax H3 that generates a five-second clip in about three seconds. It is the model the live channels we currently track are built on.
Related
Related models and reading
The phrase gets used for everything from a caption bot to a channel that paints itself into existence. This guide draws the line, explains how the real thing works, and points to channels you can watch right now.
Every directory can put a red dot next to a name. Making that dot mean something requires public evidence, expiry and a refusal to guess. This guide explains the approach AI Stream Radar uses and how you can apply it yourself.
Sources
Official sources
- Introducing H3 Max by fal
“a post-trained version of MiniMax H3 developed by fal Research and optimized for maximum speed”
Announcement post dated 27 Aug 2026. States H3 Max is post-trained from the open-weights MiniMax H3, generates a 5-second clip in roughly 3 seconds (faster than real time), and is available via fal's playground and API. Weights availability for H3 Max itself is not stated, so its open-source flag stays unknown.
namedeveloper_organization_idmodel_typerelease_datebased_on_model_slugapi_availablesupports_realtimeis_open_source - MiniMax H3 Max (Text to Video) API on fal
“a post-trained variant of MiniMax H3, tuned for stronger prompt adherence”
Model endpoint page on fal. This is the URL the fal.live homepage links to for the model.
official_url