Krafton launches a chat app that remembers you

Key Takeaways

Krafton has launched Kiff, a K-pop fandom-based AI chat app, for the first commercial use of its in-house AI technology.

Its first character collaboration is Wenny, a role developed with the girl group Fifty Fifty and described as remembering past conversations and offering a daily illustrated diary.

Some chats are free, while unlimited use costs 3,900 won a month, underscoring how game AI is now being pushed into a subscription fandom service.

Krafton has entered the market with an AI chat app called Kiff and a monthly price tag of 3,900 won.

On the surface, it looks like a place to make small talk with an idol-style character. But look closer and it is doing something more unusual: it remembers that you ate poke from a salad shop days ago, and it asks about your favorite Pepero flavors on November 11, the day Koreans often mark with the snack brand Pepero. This is no longer just an AI that answers questions. It is an AI that keeps relationships going.

That raises the real question. Is fandom-oriented conversational AI a new way to consume content, or just another platform designed to bundle emotion and convenience into a subscription fee?

Why a game company started with fandom AI

Kiff is a chat app in which users talk to characters imagined by K-pop artists. Its first partner is Wenny, an AI character created with the girl group Fifty Fifty and designed with the members' participation. Inside the app, users can chat with Wenny and view a daily illustrated diary.

The pairing looks unusual at first, but it is not just a promotional tie-in. Krafton has already been testing AI inside its games. In June, it introduced an AI-based companion called PUBG Ally in PUBG: Battlegrounds, and it also applied CPC, short for Co-Playable Character, to its life simulation title inZOI. Kiff is the company's first commercial product built around technology it developed itself.

The choice of fandom also makes strategic sense. Games are often consumed in short bursts, but fandom thrives on repeat visits. The more a favorite artist, character, or habit accumulates, the more likely users are to return. Once AI begins remembering, the service shifts from a chat utility into a place that helps hold on to emotional routines.

What a memory-based AI gives, and what it costs

The biggest differentiator in Kiff is memory. If a user mentions eating poke from a salad shop, the app may bring it up again days later. On November 11, it may ask first about Pepero preferences. It also learns details about Fifty Fifty's albums and activities, then recommends content that fits a user's interests. The conversation does not end when the screen goes dark; it continues into the next exchange.

That kind of design is plainly attractive. Ordinary chatbots answer a question and then disappear into the background. A memory-based AI, by contrast, builds context over time. Fans feel closer to the character, and the company can turn that closeness into subscription revenue. The illustrated diary works the same way. By summarizing the previous day's conversation with AI-generated artwork, it gives users the sense that they have stepped into the character's day.

But convenience can also become a burden. Memory is friendly, yet if the rules for what gets remembered are unclear, users may feel uneasy. The more private conversations, preferences, and daily habits are stored in the service, the more obvious the question becomes: how will that data be used? The article makes the basic features and pricing clear enough, but the scope of memory and how it is managed still deserve more explanation.

There is also the risk that the app pushes fandom immersion too far. Once a user is pulled into an artist's world and a character's routine, they can begin to feel less like a customer and more like part of the relationship itself. That may increase engagement and spending. The point is not that this is inherently bad, but that technical progress does not automatically guarantee a better user experience.

What this first commercial launch really means

Kiff offers some conversations for free, while a monthly subscription of 3,900 won unlocks unlimited use. Krafton says it is the first time the company has commercialized AI technology it researched in-house for general consumers. Put differently, Krafton is testing whether AI developed in the lab can become a revenue model in the real world.

The entry price looks low. For the cost of a few cups of coffee, users can spend a month chatting with an AI character, which makes the service easy to try. The real issue is not the price but the habit. The cheaper the service, the more users can become attached to the experience itself rather than the fee. Once the conversations remember you, and the app adds a diary on top, leaving becomes harder.

This move also fits Krafton's broader AI strategy. The company has been advancing its own language model, expanding AI companions inside games, and now broadening the concept into fandom chat. The goal is not just a single chatbot, but an AI ecosystem linking games, characters, and content. If it works, Krafton could move from being a game company to something closer to an AI content platform.

The next competition will be about trust

Still, technology alone will not sustain this market. The competition will depend on how naturally a character responds, how well it remembers, and how often it reaches out first. Push those things too far, however, and the experience can turn tiring. Overly persistent memory can feel invasive, and excessive recommendations can make a character seem less like a friend and more like an advertisement.

That is why Kiff's success will not depend only on the quality of its language model. It will also depend on whether Krafton handles memory carefully, avoids turning fandom intimacy into commercial fatigue, and sets boundaries that users can accept. In an age when technology can speak like a person, it also has to earn trust like one.

Krafton has now stepped outside the game world for the first time. For that step to succeed, the character must do more than talk well. Users have to feel safe coming back. The speed at which AI stores memory may matter, but the more lasting question is how politely it uses that memory.

댓글 쓰기

다음 이전