内容来源：https://www.theverge.com/entertainment/896213/nvidia-dlss-5-ai-faces-motion-smoothing

内容总结：

英伟达发布新一代图形技术DLSS 5，行业争议随之而来。该公司称其为“自2018年实时光线追踪技术问世以来最重要的突破”，可通过AI渲染提升游戏画质。然而，首批演示画面中的人物面部却呈现“AI同质化”特征——《生化危机：安魂曲》主角格蕾丝面容失真，《霍格沃茨之遗》角色如同经过滤镜处理，连真人范戴克的形象也失去辨识度。

这项定于秋季上市的技术虽为可选功能，却已获得贝塞斯达、卡普空等头部厂商支持。公司高管称赞其能“突破实时渲染的传统限制”，但演示画面引发的质疑声浪不断。批评者指出，AI生成的面部常带有皮肤过度光滑、五官趋同、表情僵化等特征，易陷入“恐怖谷效应”。

行业观察人士担忧，在游戏业经历裁员与工作室关闭浪潮的背景下，AI技术对艺术创作的介入可能加剧从业者危机。尽管部分独立开发者公开反对该技术，甚至以“无AI”作为宣传标签，但大型开发商决策层仍积极推动相关应用。目前贝塞斯达已承诺将对画面效果进行艺术调整，最终成品表现仍有待观察。

随着AI美学从社交媒体向传统娱乐领域渗透，这项技术或将引发更深层的行业讨论：当技术突破与艺术表达产生冲突时，游戏产业将如何平衡创新与创作初心？

中文翻译：

英伟达昨日公布了其最新升级技术DLSS 5，并称之为"自2018年实时光线追踪技术问世以来，公司在计算机图形领域最重大的突破"。宣传语听着不错，实际效果却令人咋舌。英伟达宣称这项技术"能为像素注入逼真的光影与材质"，但人们只注意到它把辨识度极高的面孔变成了千篇一律的AI合成脸——《生化危机：安魂曲》女主角格蕾丝经过"改造"后，简直像是蒂莉·诺伍德视频里的网红；《霍格沃茨之遗》的孩子们仿佛被强行塞进Instagram滤镜；就连现实中真实存在的利物浦队长维吉尔·范戴克，面部特征也被扭曲成了DLSS 5量产化的路人模样。

英伟达的DLSS 5好比游戏界的动态平滑技术，却更令人不安。

这种千篇一律的"AI脸"正在入侵游戏世界，而它的模样实在不敢恭维。

所谓"重大突破"给所有画面都烙上了AI生成艺术特有的印记。这有点像动态平滑技术，但更进了一步——它直接改变了人物的面部特征，让所有形象都趋于同质化。

需要说明的是，当你下次在电脑上玩《安魂曲》时，格蕾丝并不会突然变成Grok Imagine演示视频里的模样。DLSS 5要等到今年秋季才正式发布，需要高性能硬件支持，并且是可选项功能。但推动这项技术的是全球市值最高的企业之一，且获得了多家主流游戏开发商的支持。这些厂商似乎都乐于让自己的游戏沾染上这种特殊的美学风格。

在英伟达公告博客的声明中，贝塞斯达总监托德·霍华德宣称"DLSS 5能让艺术风格与细节突破传统实时渲染的限制大放异彩"，同时确认该技术将应用于《星空》游戏。卡普空多部重磅大作（包括《安魂曲》）的执行制作人竹内润则表示："DLSS 5标志着视觉保真度发展的又一重要里程碑，将帮助玩家更深度沉浸于《生化危机》的世界。"

令人费解的是，这些业界最具影响力的人物竟认为让英伟达用千篇一律的AI版本取代精心设计的角色是件很酷的事。贝塞斯达在后续推文中解释，目前展示的只是"非常早期的版本"，工作室"美术团队将继续调整光照与最终效果，为每款游戏打造最合适的视觉呈现"。或许秋季发布的DLSS 5最终版本会大有不同。

但当前呈现的景象仍指向黯淡的未来。人工智能已渗透生活的方方面面，最令人沮丧的入侵方式之一就发生在美学层面。AI生成的面孔融合了无数图像数据，最终输出的是同质化的理想模板。这种面孔通常具有若干标志性特征：非自然的柔滑肌肤与标准化五官、永远欢快的眼神、丰满嘴唇的程式化微笑、完美却虚假的发型、小巧鼻型，以及凸显轮廓的HDR式打光。单独看这些或许是常见面部特征，但当每张AI脸都具备全部或大部分特征时，我们就坠入了恐怖谷效应。

这正是英伟达演示画面引发强烈反响的原因：这些面孔不仅难看，更与其他AI产物如出一辙。从Instagram动态到YouTube缩略图，这种美学风格无处不在，正从社交网络逐渐蚕食传统娱乐文化领域。我至今未见过优秀的AI生成电影，但它们层出不穷，单看一帧画面就能辨识。英伟达的新技术正是这种美学入侵游戏领域最直观的例证。

AI扭曲艺术家作品的现象对游戏产业尤其危险。经历疫情后行业衰退与错误战略导致的裁员潮、工作室倒闭潮后，用粗制滥造的AI替代人类劳动的潜在可能令人不安。更棘手的是，部分玩家对正常女性形象存在严重认知偏差，用AI工具将现有角色变得更趋同质化、更卡通化的做法极具争议。

公告发布后，众多独立开发者通过表情包和尖锐批评表达对DLSS 5的抵制。虽然多数开发者反对在游戏中使用生成式AI，部分小型工作室甚至将"无AI"作为营销标签，但大型开发商决策层的态度截然不同。这正是霍华德与竹内润等人极力推崇该技术的原因，即便演示效果已沦为笑柄。

格蕾丝经DLSS 5渲染的面孔，预示着这项技术普及后可能出现的景象。若真如此，真正的朋友来家里做客时，你或许该像关闭动态平滑功能那样，主动关掉这个特效。

热门资讯：

• 本雅明·内塔尼亚胡竭力证明自己非AI克隆体
• 三星Galaxy Z TriFold上市仅三月即停产
• DLSS 5形同实时生成式AI游戏滤镜
• 微软AI管理层重组后任命新任Copilot主管
• 探访可能重塑迪士尼乐园未来的"奥拉夫" Frozen机器人

英文来源：

Yesterday Nvidia revealed its latest upscaling tech, called DLSS 5, which it described as “the company’s most significant breakthrough in computer graphics since the debut of real-time ray tracing in 2018.” Sounds good, until you actually see it. According to Nvidia, the tech “infuses pixels with photoreal lighting and materials,” but all anyone seemed to notice was that it turned recognizable faces into something resembling AI slop. Resident Evil Requiem protagonist Grace got a makeover that would make her look at home in a Tilly Norwood video. The Hogwarts Legacy kids looked like they’d been wrung through an Instagram filter. Even Liverpool captain Virgil van Dijk, a very real and famous person, had his features warped and became just some other dude with DLSS 5.
Nvidia’s DLSS 5 is like motion smoothing for video games, but worse
The generic, homogenized ‘AI face’ has come for gaming, and it doesn’t look good.
The generic, homogenized ‘AI face’ has come for gaming, and it doesn’t look good.
This “significant breakthrough” imbues everything with a particular look that’s become synonymous with AI-generated art. It’s sort of like motion smoothing, if motion smoothing went a step further and changed people’s faces — and it’s making everything look the same.
It’s important to note that the next time you play Requiem on a PC, Grace won’t suddenly look like she was ripped out of a Grok Imagine demo. DLSS 5 doesn’t launch until the fall, it’ll require some beefy hardware to operate, and it is an optional feature. But it is a technology that is being pushed by one of the most valuable companies in the world, which has support from major video game developers. And they all seem content with associating their games with a very particular aesthetic.
In a statement on Nvidia’s announcement blog, Bethesda boss Todd Howard said that “with DLSS 5 the artistic style and detail shine through without being held back by the traditional limits of real-time rendering,” while noting that the feature will be available in Starfield. Jun Takeuchi, executive producer on many of Capcom’s biggest blockbusters, including Requiem, said that “DLSS 5 represents another important step in pushing visual fidelity forward, helping players become even more immersed in the world of Resident Evil.”
It’s a little strange to hear that some of the most influential names in games have decided that it’s cool for Nvidia to replace their carefully crafted characters with generic AI-powered versions. In a follow-up tweet, Bethesda noted that what we’re seeing is a “very early look,” and that the studio’s “art teams will be further adjusting the lighting and final effect to look the way we think works best for each game.” So maybe the version of DLSS 5 that’s available in the fall will look very different.
But what we are seeing now points to a bleak future. AI has infiltrated nearly every aspect of our lives, and one of the most frustrating ways has been on an aesthetic level. AI-generated faces are an amalgamation of countless images, which are then used to spit out a sort of homogenized ideal. It’s typically easy to identify thanks to a handful of telltale signs: unnaturally smooth skin and uniform features, perpetually cheerful eyes, a smiling mouth with full lips, perfectly styled hair that looks synthetic, small noses, and HDR-style lighting that highlights every contour. On their own, these can be typical facial features, but when every AI face has them all or most of them, we start veering into the uncanny valley.
That’s why so many people reacted strongly to the faces in Nvidia’s announcement: They don’t just look bad, they look the same as everything else. That same aesthetic is prevalent everywhere from Instagram feeds to YouTube thumbnails, and it’s been inching its way from social networks to more traditional forms of entertainment and culture. I’ve yet to see a good AI-generated film, and yet they keep coming, and you can identify them from a single screen. Nvidia’s new tech is the most visible example of that aesthetic infiltrating games.
There are a number of reasons why seeing AI mangle an artist’s work is troublesome for games in particular. The industry has been ravaged by layoffs and studio closures following some very expensive misplaced bets and a post-pandemic slowdown, so the potential for replacing human work with slop doesn’t sit well. It’s also a medium where a subset of the audience has some very backward ideas about what a normal human woman looks like, so making existing characters somehow both more generic and more cartoonish through an AI tool is extremely problematic.
Since the announcement, there have been a number of indie developers who have come out to lambast DLSS 5 through memes and more explicitly negative statements. And while a large percentage of developers seem to be against the idea of using generative AI in their games, with some smaller studios going so far as to use “AI free” as a marketing label, it’s also true that many of those in decision-making positions at larger developers don’t feel the same way. That’s why we have the likes of Howard and Takeuchi espousing this technology while at the same time showing a demo that people can’t stop making fun of.
Grace’s face rendered through DLSS 5 is an early vision of what things could look like if the adoption becomes more widespread. And if that happens, being a good friend might mean turning it off when you visit, just like motion smoothing.
Most Popular

• Benjamin Netanyahu is struggling to prove he’s not an AI clone
• Samsung discontinues its Galaxy Z TriFold after just three months
• DLSS 5 looks like a real-time generative AI filter for video games
• Microsoft appoints a new Copilot boss after AI leadership shake-up
• I met Olaf — the Frozen robot who might be the future of Disney Parks