South Korean AI Company Wrtn Transforms Free Writing Tool into AI Narrative Business
South Korean AI company Wrtn Technologies has shifted its initial AI writing and chat tool towards interactive storytelling application Crack and character dialogue business, aiming for approximately $70 million in annualized revenue by the end of 2025. The company states that AI entertainment products like Crack in Korea and Kyarapu in Japan are driving rapid revenue growth, with a target to exceed $100 million in annualized revenue by 2026, and plans to enter the U.S. market by mid-2026, with an IPO potentially starting as early as 2028.
Wrtn's core product logic is not about having users read preset scripts or merely clicking options A, B, or C, but rather allowing users to act directly in natural language. Users can set identities, appearances, character relationships, and story worlds, such as campus, romance, martial arts, or magic academy; the AI continuously generates subsequent plots based on user input, while also playing other roles, remembering previous plot points, and maintaining narrative continuity through character relationships, scene changes, and emotional developments. Users are not consuming a fixed piece of content, but rather a personal storyline shaped by their participation that cannot be fully replicated by others.
Founded in 2021, Wrtn initially focused on chatbot aggregation and AI writing services, later integrating large models and quickly acquiring users through a free usage strategy. The company had previously opened access to models like GPT-4 for Korean users for free, exchanging high inference costs for user scale, usage habits, and dialogue data; this strategy helped Wrtn establish a user base of millions in Korea, but the free assistant itself did not constitute stable revenue, with the real commercial pivot coming from monetizing character dialogue and interactive narratives.
After launching the Korean version of Crack at the end of 2024, the company will operate character chat and continuous plots as independent consumable products; the Japanese market will use the Kyarapu brand, targeting users familiar with anime, games, and role-playing culture. Wrtn has not replicated the same product under a unified brand globally, but rather segmented its positioning according to regional narrative preferences: Korea emphasizes Crack's interactive entertainment experience, Japan focuses on Kyarapu's character relationships and two-dimensional context, while the U.S. is preparing to enter the role-playing user group under the OOC brand.
Wrtn's payment mechanism is designed around "generating the next segment of the plot." Users can purchase character responses, story progression, and high-frequency interactions through in-app currency, rather than just paying a monthly subscription fee; the company previously disclosed that its character chat service achieved approximately 1 billion KRW in monthly revenue in November 2024, rising to about 2 billion KRW in December. The combination of pay-per-use and plot continuity means that revenue depends not only on whether users subscribe, but also on whether they continue to advance the story and are willing to pay for longer character interactions and higher-frequency generation.
As of early 2026, Wrtn's total monthly active users in Korea and Japan exceeded 5 million. The company predicts that with the expansion of its AI entertainment business, annualized revenue will reach $700 million by the end of 2027, ten times the $70 million level at the end of 2025; this goal relies on sustained growth in the Japanese market, successful entry into the U.S. market, deeper user payment engagement, and continued reduction in model inference costs. It is important to note that the "$70 million annualized revenue" disclosed by Wrtn does not fully align with the approximately 47.1 billion KRW reported by Korean media for actual annual revenue in 2025: the former is closer to the year-end revenue run rate, while the latter is the revenue recognized in the accounting year and cannot be directly compared.
In terms of market mechanisms, young users, role-playing enthusiasts, anime and game audiences, and online users with strong feelings of loneliness are Wrtn's buyers; what Wrtn sells is continuous generation, personalized memory, and emotional engagement. Large model companies provide inference capabilities and charge based on usage costs, while Wrtn packages general models into consumable character relationships and story progress; app stores, payment channels, and creator communities extract distribution or transaction value from this. Beneficiaries include AI entertainment applications that master user relationships, character assets, and payment design; those under pressure are general AI tools that only provide one-time writing generation and lack long-term interactive scenarios.
Source: Public Information
ABAB AI Insight
Wrtn's historical path reflects a typical transformation of South Korean consumer AI companies from "model entry" to "entertainment entry." When it was founded in 2021, the company primarily focused on chatbot aggregation and productivity tools, aiming to make it easier for users to utilize different large models; as general assistants like ChatGPT, Claude, and Gemini continuously lower the usage threshold, merely providing model entry becomes difficult to maintain differentiation. Wrtn did not compete with OpenAI or Google over "whose model is stronger," but embedded model capabilities into character relationships, continuous plots, and interactive entertainment, shifting the competitive dimension from answer accuracy to user engagement time and emotional investment.
The free GPT strategy was key to its early capital path. Large model invocation costs are high, and during the free phase, Wrtn effectively subsidized user inference costs with investment funds in exchange for scale, habits, and brand recognition; this is not fundamentally different from how early internet companies used subsidies to acquire users. The difference is that the marginal cost of AI products is not close to zero: each generation incurs model inference fees. Therefore, Wrtn must subsequently convert "free Q&A users" into "entertainment users willing to continuously pay for the next segment of the story"; otherwise, the larger the user scale, the greater the pressure from inference costs.
The commercial value of Crack and Kyarapu comes from "narrative memory." Users of ordinary AI writing tools can easily migrate to another model, continuing their creation after copying prompts; however, interactive storytelling applications remember users' identities, relationships, character preferences, historical choices, and plot branches. Leaving the application does not just mean switching chat windows, but interrupting a personal narrative that has already invested time and emotion. This migration cost does not stem from technical lock-in, but from users' sunk investment in virtual relationships and continuous stories.
A historical analogy is the paid structure of South Korean web novels, web comics, and Japanese mobile games. Web novels typically charge for chapter unlocks, while mobile games monetize through card draws, stamina, skins, and story progression; Wrtn has transplanted these mature content business models into generative AI. Users do not purchase a completed novel, but rather the "next segment generated just for themselves"; users do not buy a fixed character, but rather the ability for the character to continuously respond, remember them, and participate in a shared world.
Wrtn employs different brands in Korea, Japan, and the U.S., essentially localizing narrative products rather than simply translating. The character chat in the Korean market is closer to social entertainment and real-time interaction, while the Japanese market emphasizes anime, games, and character culture, and the U.S. market's OOC directly borrows from role-playing community language. Models can be deployed across languages, but the types of fantasies users are willing to pay for, relationship boundaries, expression habits, and content aesthetics are highly localized; this is also why Wrtn needs to adjust models, character templates, and community operations for different markets.
This represents a transfer of pricing power. After general models rapidly commoditize "text generation," text itself is no longer scarce; platforms that can continuously occupy user time, organize character assets, retain story memories, and design payment nodes will hold pricing power. The future competition in AI consumer applications will not be about who first accesses the latest models, but about who can turn model outputs into a behavior loop that users are willing to repeatedly purchase.
ABAB News · Cognitive Laws
Model generates text, story generates repurchase
Free tools grab users, continuous relationships grab revenue
Content can be infinitely generated, but memory can form a moat