Feasibility of using Artist-AI co-creativity workflows for remastering classic video games

使用 Artist-AI 共同创意工作流程重新制作经典视频游戏的可行性

• 批准号: 10081575
• 金额: $ 5.94万
• 依托单位: PIXEL RESEARCH LTD
• 依托单位国家: 英国
• 项目类别: Collaborative R&D
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=10081575
• 关键词: Feasibility; using; Artist; AI; co

In the video games industry, there is currently the opportunity to remaster classic game titles from previous generations of gaming. This trend has gained momentum in recent years, with many game studios and publishers remastering their classic games for modern platforms. Remastering older titles can be an excellent way for game studios to generate substantial additional revenue by appealing to both older fans and new audiences who may not have played the original and to preserve classic titles and keep them relevant for a new generation of gamers. Remastering typically involves enhancing the graphics and sound of the original game and improving the user experience to make it more appealing and accessible to modern audiences. There is however still a large time commitment and the upfront cost of remastering classic game titles that can be prohibitively expensive.This project will investigate the feasibility of implementing Artificial Intelligence (AI), to integrate into existing workflows, in order to reduce the overall work hours, and therefore upfront cost, required for the development of remastered versions of classic games. Specifically, we will investigate how AI can be developed, trained and utilised to upscale very low-resolution images, in order to remaster classic game titles. Traditionally, image upscaling involved simple methods such as interpolation, which would create new pixels by estimating the values between the existing pixels. However, these methods often resulted in blurry or pixelated images. In recent years, there has been a significant advancement in image upscaling technology using AI. AI can learn to identify patterns in low-resolution images and generate high-resolution versions of the same image with significantly improved quality.Whilst there are a number of companies and research institutions investigating upscaling technology, there is currently no investigation, product or research into how to upscale very-low-resolution hand-drawn art assets for game asset remastering. This project will investigate the feasibility of integrating current upscaling technologies within this space into existing game studio workflows and develop a generalised, commercialisable tool, as a demonstrator.

在视频游戏行业，目前有机会重新掌握前几代游戏的经典游戏标题。近年来，这一趋势越来越明显，许多游戏工作室和发行商都在为现代平台重新制作经典游戏。重制旧游戏是游戏工作室创造大量额外收入的绝佳方式，既可以吸引老粉丝，也可以吸引可能没有玩过原版游戏的新观众，还可以保留经典游戏，让它们与新一代游戏玩家保持相关性。重制通常涉及增强原始游戏的图形和声音，并改善用户体验，使其更吸引现代观众。然而，重制经典游戏仍然需要投入大量的时间和高昂的前期成本。本项目将研究将人工智能（AI）整合到现有工作流程中的可行性，以减少开发重制版经典游戏所需的总工作时间和前期成本。具体来说，我们将研究如何开发、训练和利用人工智能来提升极低分辨率的图像，以重新掌握经典游戏。传统上，图像放大涉及简单的方法，如插值，这将通过估计现有像素之间的值来创建新的像素。然而，这些方法通常导致模糊或像素化的图像。近年来，使用AI的图像放大技术取得了重大进展。人工智能可以学习识别低分辨率图像中的模式，并生成相同图像的高分辨率版本，并显着提高质量。虽然有一些公司和研究机构正在研究升级技术，但目前还没有调查，产品或研究如何升级极低分辨率的手绘艺术资产用于游戏资产重制。该项目将研究将当前的升级技术整合到现有游戏工作室工作流程中的可行性，并开发一个通用的、可商业化的工具作为演示。