What is Happy Oyster?
Happy Oyster is Alibaba's real-time world model and, alongside Genie, one of the few closed models in this category that already feels like a product rather than a lab demo. It does not hand back a finished clip: it generates the world while you steer it, taking text, voice, image and keyboard input and producing picture and sound together in one continuous 720p, 24 FPS stream.
It comes from the same Alibaba team behind HappyHorse-1.0, the video model that topped the Artificial Analysis video leaderboards before Alibaba confirmed it owned it.
Two ways to play: Directing and Wandering
Directing lets you steer a running scene for up to three minutes: change camera angles, redirect characters or alter the story while it is being generated. Wandering drops you into first person with standard WASD and camera controls, and the world keeps extending as you walk, for up to a minute per session.
Why the audio matters
Most world models treat sound as post-production, if they have it at all. Happy Oyster generates visual and auditory signals inside the same world state, so footsteps, weather and voices are aligned with what you see by construction rather than synchronised afterwards. Once you are inside a scene, that small difference matters a lot for immersion.
Strengths, weaknesses and access
Independent benchmarking puts Happy Oyster near the top for subject memory in third person and for interaction physics in first person: characters stay themselves and objects behave. Its weaker area is action-following, where the strongest open models still beat it. The session limits also place it closer to prototyping and storyboarding than to finished production.
Access is through an early-access waitlist on the official site, and developers can call it through fal's API.
Strengths and limitations
Strengths
- Picture and sound generated together in real time
- Text, voice, image and keyboard input
- Strong subject memory and interaction physics in benchmarks
- Developer access through fal
Limitations
- Closed model with waitlisted access
- Short sessions: about 1 minute (Wandering) to 3 minutes (Directing)
- Weaker at following specific actions than the best open models
Who is Happy Oyster for?
Creators and studios who want a polished, real-time world with sound for prototyping, storyboarding and interactive previews, and developers who prefer calling an API over running models themselves.
Frequently asked questions
Who makes Happy Oyster?
Happy Oyster is developed by Alibaba, through Alibaba Token Hub — the same team behind the HappyHorse-1.0 video model.
Does Happy Oyster generate sound?
Yes. It generates audio and video together inside the same world state, so sound and picture stay aligned.
How long can a Happy Oyster session last?
Directing mode lets you steer a scene for up to three minutes; Wandering mode, in first person, lasts up to a minute per session.
Is Happy Oyster open source?
No. It is a closed model available through early access and fal's API.
How do I try Happy Oyster?
Join the waitlist on the official Happy Oyster site, or use it through fal if you are a developer.