Kuaikan Launches AI Interactive Story Product Livo, Tested as an 'AI Westworld'

Over the past few years, the path by which AI entered narrative has roughly been: first let models continue plots, then give characters memory, goals, and social interaction, and finally try to let the world continue evolving after the user leaves. The thread traced by the source includes Latitude's AI Dungeon in 2019, Stanford University and Google's Generative Agents in 2023 (25 Agents), Unbounded's "infinite game" in 2024, Altera's Project Sid from the same period, which placed more than 1,000 Agents into Minecraft, and Voyage, launched by Latitude in 2026 with a World Engine added.

But the source raises a key distinction: a world that can run is not the same as a good story. Agents can chat, act, and even form societies, but how suspense is laid out and how relationships generate tension still depends on content design. This is precisely the background against which Kuaikan chose to start from content and launch Livo.

Livo is compared by the source to an "AI tavern": creators first build the worldview, main characters, and key plot, and then multiple Agents maintain the characters' emotions, memories, relationships, and daily actions. Users receive an identity within a world and enter a story in which conflicts, secrets, and character fates have already been planted. Four premium story worlds have been released so far, covering campus, urban, espionage, and fantasy themes.

The source takes "The Winter of Tilia" as an example: the story is set on a transcontinental train called the "Peace," where the user boards with the XK-101 sample contested by all sides and is drawn into a contest among multiple factions. The opening animation explains the background of the war and the identity setup. From the very beginning, the user has a clear origin, mission, and secret, but who is trustworthy and how much information others hold must be gradually judged within the story.

What then appears is a top-down train map, with character avatars distributed across locations such as the reading room, warehouse, and second-class carriage, each acting according to their own schedule; below the map, world dynamics are continuously updated, recording new characters appearing, character actions, and events currently taking place. According to the source, even without user intervention, this world is running on its own.

The story has two ways of advancing: following the "destined fate" mainline set by the creator, or entering a temporarily generated "potential fate" based on character states and current relationships; "fate that has already occurred" preserves the plot the user has personally experienced. The interface displays the character's current emotions, and Deep Feeling can show how the character understands what the user has just said, with the system presenting their state along dimensions such as weighing and savoring, intimacy and distance, and tenderness and killing intent.

The source points out that it is not difficult for a character to maintain their persona over one or two rounds of dialogue; the real test is whether they can still seem like themselves after interaction is stretched out: if a character never changes at all, they lose freshness, but if changes lack a basis, personality drift appears. A cautious and suspicious person cannot suddenly deliver trust because of a few friendly words.

For this reason, Livo gives characters two sets of states: a six-layer personality profile preserves the relatively stable parts, including self-awareness, underlying needs, beliefs, ways of thinking, stance, and emotional tendencies; real-time changing emotional states record the character's feelings in specific situations. New experiences affect current judgment, and important experiences settle into long-term relationships.

Character changes are also recorded. The "soul map" forms a personality outline from traits such as clear-headed and rational, skilled at disguise, ambitious, and cold-blooded and cruel. Some information is initially hidden and only gradually revealed as communication deepens; "soul evolution" records the reasons changes occur. "See me" records the character's understanding of the user, the user's knowledge of the character, and the influence the user has had on them.

Openness brings another challenge: traditional interactive narrative relies on branching trees, where every option and result is written in advance by the creator, and the more complex the content, the higher the cost; if it is entirely handed over to AI generation, it easily loses rhythm and may even deviate from character motivations and mainline goals.

Livo separates the two parts: creators determine world rules, core conflicts, and key turning points; the mainline defines the story's direction; and manually written "red situations" carry important scenes; AI-generated "black situations" are responsible for connecting these nodes and supplementing personalized side branches according to character states. The three types of "fate" mentioned earlier are the presentation of this division of labor in the product.

After structuring is completed, different Agents begin to divide the work: the Soul Agent is responsible for the character's judgments and actions, the World Agent schedules daily routines, events, environments, and relationships, Deep Feeling handles the emotional chain from perception to expression, and the Scribe Agent supplements situations within the boundaries set by the creator. These Agents also need to share memory and causality: Livo stores information according to emotional importance, ordinary pleasantries may quickly fade, while conversations involving secrets, promises, or betrayal receive higher weight; memories between different story worlds are isolated from one another; the long-range causality system connects key results, stage events, and character choices to avoid obvious contradictions after the story has been running for a long time. The source therefore judges that Livo's technical difficulty lies not only in model capability, but more crucially in translating literary language into psychological structure, and then turning psychological structure into data that can continue to run.

The source is clear about Livo's positioning: from the perspective of actual experience, Livo at the current stage is still an early exploratory product. Using AI to build a living world is itself a complex project and faces many difficult problems that need to be overcome.

The source also proposes that over the past two years, the combination of AI and the content industry has mostly centered on increasing production capacity—generating images, videos, and scripts to make content faster and cheaper; Livo's vision, by contrast, is to let AI enter the interior of the story and participate in the long-term operation of characters and the world. Traditional works are usually completed before release, with plot, character arcs, and endings already determined; after entering this kind of AI-native product, release is closer to an opening, where creators provide world rules, character personalities, and key plot, and user participation continuously adds new conditions.

This also changes creators' responsibilities and the dimensions by which work quality is measured: in addition to writing attractive characters and conflicts, creators must also design how characters judge, the boundaries of their actions, and the conditions for change; whether characters can remain consistent over the long term, whether generated content can maintain causality, and whether user choices can leave a real impact will all determine how far the experience can go. The source believes that how to tell stories in the AI era has become an unavoidable topic for entertainment industry practitioners, and Kuaikan has taken the lead in offering its own solution; this may be the greatest industry significance of Livo's launch.