Skip to content

Fields

NPCs & dialogue latest news

Dialogue, memory, behavior and story for generative NPCs: middleware, shipped games, and the engineering of latency and cost.

38 picks19 in the last 30 days64 collected in all

Updated

Picks archive · Page 2

Oct 21Wed21–38
  1. arXiv Game AI42

    WorldStream: streaming interactive 3D worlds with hierarchical planning and persistent state

    WorldStream introduces an open-ended 3D world model that can expand playable terrain at runtime, respond to open-ended language instructions, and restore visited locations along with their edits. A hierarchical planner outputs macro and micro structural decisions, which deterministic local code then compiles into terrain, roads, settlements, vegetation, and entities. The world record stores plans, seeds, object identities, mutable state, and boundary constraints, so returning chunks can be recompiled from it. Stateful NPCs also update as they act.

    Why it matters: The paper puts terrain generation, NPC state, and records of edits to visited areas into one shared world state. It is useful for comparing approaches to streaming open-world generation.

Oct 19Mon
  1. Youxituoluo46

    How much would it cost to have game NPCs remember players?

    The article assumes 100,000 monthly active players with 100 dialogue rounds per player per month, and uses public large model API prices to estimate that model inference and memory processing would cost about US$6,908 per month. This excludes costs such as voice, servers, and operations. The article notes that NPC memory should be split into recent raw text, summaries of events and preferences, and on-demand retrieval, and that costs grow roughly linearly with player numbers.

    Why it matters: The article breaks down costs based on an assumption of 100,000 monthly active players with 100 dialogue rounds per player each month, and compares the costs of cloud APIs with those of on-device deployment. It can serve as a reference for estimating your own project.

Oct 16Fri
  1. Youxituoluo44

    Character.AI bankruptcy: The cost and retention dilemma of AI character chat

    Character.AI has filed for bankruptcy, with a deal expected to close in late November or early December; if no buyer willing to keep the service running is found, the platform may gradually shut down. Bankruptcy filings show its AI computing costs peaked at $13.8 million per month in 2024 and were still around $10 million per month in 2025, while subscription revenue peaked at just $2.8 million. The article also notes that conversations failing to build into an experience users keep accumulating are a major challenge for retention and monetization.

    Why it matters: Using Character.AI's bankruptcy as a case study, this piece breaks down the inference costs and user retention challenges facing AI character chat products, offering useful lessons for teams building AI character gameplay.

Oct 15Thu
  1. GameLook41

    Kurogames Makes First AI Investment in Yanzhi Domain; Gaming Industry's AI Investment Momentum Continues

    According to company registration data, Kurogames invested in Yanzhi Domain Technology (Shenzhen) Co., Ltd., holding a 15% stake—Kurogames' first AI company investment. Founded in 2024, Yanzhi Domain's products include Vocu, an AI audio creation platform, and Genyu TalkiVerse, an interactive 3D AI character platform for scenarios like game NPCs. The article also notes that miHoYo, Qihoo, and Giant Network have increased AI investments in recent years across large models and AIGC.

    Why it matters: The article contextualizes Kurogames' investment within Tencent's holdings and continued industry bets on AI by gacha publishers, making it a good reference for understanding the broader trajectory of Chinese gaming companies' AI investments.

Oct 14Wed
  1. Inworld AI Blog49

    Inworld launches Inference service for consumer AI

    Inworld has launched Inworld Inference, an inference service built for consumer AI applications and now open to all developers. The company says that on p95 time to first token, Gemma 4 26B is 7.8 times faster than the median OpenRouter provider, with input priced at $0.07 per 1M input tokens.

    Why it matters: The article breaks down the latency and cost drivers of a conversational AI turn, from time to first token and caching to routing, and can be used to compare inference service options for conversational AI.

  2. GameLook47

    Yu Huasong of Yoozoo on how AI agents are reshaping game creation and interaction

    Yu Huasong, head of the Hong'an Lab at Yoozoo Games' AI Innovation Institute, shared at the 2026 China Metaverse Conference that AI agents are pushing AI from a game production tool toward in-game agents and builders of game worlds. The talk noted that six major scenarios and more than 100 sub-scenarios inside Yoozoo have achieved full AI coverage, and described the perception, memory, goals and decision-making structure of intelligent NPCs. The decision-making large model handles open-ended decisions, while deterministic rules such as damage calculation are still handled by the game engine.

    Why it matters: The piece separates the intelligent NPC's decision-making large model from the engine's deterministic rules, which is useful for comparing against your own project: which logic should be handed to the model and which should still be computed by the engine.

Oct 13Tue
  1. GamesIndustry.biz38

    Dragon Quest creator Yuji Horii releases game he says he made with AI in about 30 minutes

    Dragon Quest creator Yuji Horii has released a text-based interrogation game on X that he says was made with AI and took about 30 minutes to build, and he warns players not to press the speed-up button. He says that with AI, even people who can't draw or program can make a decent game on their own just by giving instructions, and he compares this with his experience of completing The Portopia Serial Murder Case by himself in 1983.

    Why it matters: Yuji Horii made a text-based interrogation game with AI in about 30 minutes, which can be set against his experience of completing The Portopia Serial Murder Case on his own in 1983.

Oct 12Mon
  1. GameLook60

    Jiao Meimei Kai Che demo goes viral as developer survives on bank loans with 7,000 yuan daily AI bill

    The driving-instructor simulation game Jiao Meimei Kai Che, developed by the four-person indie team Easy Fox, has gone viral on Steam and social platforms, with playtest players growing more than 20-fold in a month. Because the sister's voice is generated in real time by a cloud large language model, daily AI service costs have exceeded 160,000 yen (about RMB 7,000), leaving the team's finances on the verge of collapse; it has applied for a bank loan to keep its servers running and admits the demo may be taken offline early.

    Why it matters: The indie team was dragged into debt by cloud large-model token costs when players flooded in, exposing the cost-structure problem that real-time AI dialogue gameplay faces on the business side.

Oct 8Thu
  1. YYS (Youyanshe)43

    Arena Breakout S19 adds the card mode Arena Breakout: Blockade Line and launches Commander Mode in a Girls' Frontline 2 crossover

    Since Season 19 went live on September 2, Arena Breakout has rolled out Hardcore Mode 2.0, Commander Mode and the card mode Arena Breakout: Blockade Line, and on September 30 it launched a crossover with Girls' Frontline 2. The card mode is turn-based deck building, with cards split into four types — commander, soldier, weapon and order — three fronts on the field, and 5 preset starter decks. Commander Mode lets players command a squad of AI soldiers with a built-in wheel or voice, and the crossover characters OTs-14 and Klukai can be added to the squad.

    Why it matters: Arena Breakout is running two side tracks — cards and AI command — alongside its core extraction gameplay, making it a sample case for how the shooter genre expands its audience.

Oct 2Fri
Sep 30Wed
  1. Youxituoluo43

    Vidu S2 Powers AI Streamer Ziying's Mid-Autumn Broadcast, Drawing Over 300,000 Viewers

    During the Mid-Autumn Festival, Zaomeng Ciyuan's AI streamer Ziying took on a time-limited broadcast challenge in her live room; the real-time interaction model behind her is Vidu S2, released not long before, and the room drew as many as 20,000 viewers at one point. Ziying has no fixed script: viewers can request songs, vote on her look and send gifts, and she has to change outfits and perform while interacting, then pick up where she left off after being interrupted. The author also tested Vidu S2's Avatar and Editing: uploading a real person's photo or an anime image and adding some settings creates a digital human, and switching on the camera allows real-time outfit changes, art style changes and background swaps while filming.

    Why it matters: Using a livestream hosted by an AI streamer with no fixed script as a case study, it tests Vidu S2's real-time interaction and outfit-switching abilities, and discusses how they could be applied to game NPCs.

  2. GameLook48

    Tencent quietly closed-beta tests E Ciyuan, an AI game companion product built around digital human buddies and voice companionship

    Tencent is quietly running a closed beta of E Ciyuan, an AI game companion product that offers digital human characters, voice chat and real-time screen recognition. Players can pick male or female companions with different personas for everyday conversation, or turn on a game companion mode. The product is free during the beta, and the interface already shows a framework for exchanging star coins for energy and for spending energy by conversation turn, voice or length of companionship; the launch date, the list of supported titles and the final pricing plan have not been made public.

    Why it matters: The article places E Ciyuan alongside in-game companion examples such as Lingbao and AI Pom-Pom, showing the difficulties standalone AI companions face in adaptation and monetization.

  3. Inworld AI Blog51

    Inworld acquires realtime voice agent platform Ultravox

    Inworld has announced the acquisition of Ultravox, a realtime voice agent platform; the Ultravox team has joined Inworld and continues to develop the platform. The first update after the acquisition is that the built-in Inworld voice in Ultravox now uses Realtime TTS-2, existing voice IDs remain usable with no code changes, and the upgrade costs nothing extra.

    Why it matters: The acquisition brings voice understanding, reasoning and voice generation into one realtime inference infrastructure, letting readers judge how the voice agent stack is consolidating.

Sep 29Tue
  1. Naavik52

    From Disney to miHoYo: the road toward AI-enabled fandom

    The article argues that IP holders are shifting from suppressing unlicensed AI characters to promoting official, controllable AI character interactions. Disney's appointment of Karandeep Anand of Character.AI as its first company-wide CTO is cited as a signal.

    Why it matters: This article compares approaches from Disney, miHoYo, Hasbro and others, useful for seeing how IP holders split control over character behavior between licensing and their own channels.

Aug 31Mon
  1. Inworld AI Blog57

    Inworld AI releases Realtime TTS-2 realtime conversational voice model

    Inworld AI has released Realtime TTS-2, a new-generation realtime conversational voice model, now fully available in the Inworld API and the Inworld Realtime API.

    Why it matters: The original post covers multi-turn audio context, natural-language voice instructions and a first-audio latency of under 200ms, which readers can use to judge whether realtime voice conversation fits their own projects.

Aug 25Tue
  1. NVIDIA Blog38

    EA, Ubisoft, Embark and other publishers bring PC games and anticheat support to NVIDIA RTX Spark

    NVIDIA announced at Gamescom that publishers including EA, Embark and Ubisoft will bring games and anticheat support to RTX Spark, which launches this fall, with EA Javelin Anticheat also natively supporting the platform.

    Why it matters: The partner list ahead of RTX Spark's fall launch grows from May's COMPUTEX to this Gamescom, so readers can compare the two batches of publishers to gauge how adoption of the new platform is progressing.

Jun 15Mon
  1. Inworld AI Blog38

    Building conversational voice agents with Mastra and Inworld speech-to-speech

    Inworld Realtime API links STT, LLM, and TTS into a single WebSocket session, and the new Mastra package @mastra/voice-inworld-realtime can bind that session as an agent's voice.

    Why it matters: The article shows how to bring STT, LLM, and TTS into a single WebSocket session, and its approach to handling interruptions and tool calls can be compared with your own voice agent setup.