Sign in
Pending benchmarksTracking since Oct 30, 2026

Is HappyOyster Acting state of the art?Pending benchmarks.Announced Sep 17, 2026; no independent results yet.

Alibaba's real-time interactive role-play model, the Acting mode of its HappyOyster world-model series: it builds a character from a text prompt and one first-frame image, then streams the character's video and speech live while you steer its lines, expressions and actions with text. API only, on Alibaba Cloud Model Studio (invitational preview) and the Reactor API.

Announced facts

The details so far

Lab
Alibaba / Qwen
Announced
Sep 17, 2026
API
Live · happyoyster-1.0-acting · Alibaba Cloud Model Studio as happyoyster-1.0-acting (Singapore and US (Virginia)), invitational preview; Reactor API as reactor/happy-oyster-acting
Open weights
Unknown · None announced
Type
Real-time interactive role-play (character) model; Alibaba lists it under World models
Family
HappyOyster 1.0, alongside the Adventure and Directing modes; each mode is deployed separately with its own API
Input
A text prompt (up to 2,000 characters) plus one required first-frame image
Output
Live video stream of the character with generated speech
Interaction
Text instructions during the session; pause and resume, no rewind
Resolution
480p (default) or 720p; 9:16 portrait (default) or 16:9
Session length
About 3 minutes of active playback per performance
Release
Alibaba Cloud Model Studio Sep 17, 2026 (model lifecycle page); Reactor API Oct 30, 2026

Tags marked unsure are not stated on the lab's own pages.

Next steps

What would let us call it

Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.

  • Publicly announced with a source
  • Usable by the public (API or weights)
  • Listed on an independent board we track
  • Enough results for a rank in a category
  • Holds across two snapshots (crowns only)
Self-reported

What the lab claims

Self-reported. Shown for context; not used in any verdict.

Lab claim: the character responds in real time with expressions, actions and voice, and its appearance, voice and personality stay consistent for 3 minutes.

Lab claim: Reactor's launch post says a character can be created from any image, with its dialogue, expressions and gestures controlled by text.

Receipts

Sources

LABAlibaba Cloud Model Studio: model lifecycle and updatesalibabacloud.com/help/en/model-studio/newly-released-models · Sep 17, 2026LABAlibaba Cloud Model Studio: happyoyster-1.0-acting model infoalibabacloud.com/help/en/model-studio/happyoyster-1-0-acting · Oct 11, 2026LABAlibaba Cloud Model Studio: HappyOyster user guidealibabacloud.com/help/en/model-studio/happyoyster-guide · Oct 11, 2026PRESSReactor announcement on Xx.com/reactorworld/status/2116231219319132231 · Oct 30, 2026PRESSReactor API docs: HappyOysterdocs.reactor.inc/model-api-reference/happy-oyster/overview · Oct 30, 2026

Tracking continues until an independent board lists it.