76万点赞，降到5万。用了不到两个月。
AI选秀节目《12星练赛》，抖音账号粉丝53.9万，初舞台评级视频点赞76.5万，先导片73.4万——刚上线的时候，那叫一个火爆。AI偶像、12星座人设、选秀赛制……每一个点都踩中了流量密码。
但热度退得也快。到了8月的夏日运动会系列视频，点赞量只有7.5万和5.3万——不到峰值的十分之一。
从76万到5万，这个跌幅背后藏着一个很多人不愿面对的事实：AI可以复制一切形式，但复制不了灵魂。
为什么选秀节目非得有"真人"不可？
很多人可能觉得，选秀嘛，不就是看唱歌跳舞表演？AI生成的偶像唱得更好、跳得更齐、长得更完美，为什么观众不买账？
因为选秀节目的核心魅力，从来不是表演本身。
选秀真正让人上头的，是「成长感」——一个普通人从素人到明星的蜕变过程，是你看着她/他从笨拙到熟练、从紧张到从容、从自卑到自信的每一步。你投票、你打call、你熬夜拉票，本质上是在参与一个人的成长。
还有「真实感」——真人会出错、会忘词、会哭、会笑、会在镜头前说不出话。这些"不完美"恰恰是最打动人的地方。AI偶像永远不会出错，永远完美，但完美到一定程度，就假了。
以及最重要的「人格魅力」——你喜欢一个选秀选手，往往不是因为她唱得最好，而是因为她的性格、她的故事、她说话的方式、她笑起来的样子。这些东西是独一无二的，是一个人十几年人生经历积累出来的。
AI偶像是一张白纸。你可以给它设定12星座的人设，可以写好每一句台词，可以设计好每一个表情——但这些都是「设计出来的性格」，不是真实的人格。观众看几期觉得新鲜，看多了就会发现：这些人设再精致，本质上和一个游戏角色没什么区别。
"你会为一个游戏角色的成长真情实感地哭吗？不会。新鲜感过了之后，你也不会为AI偶像的选秀真情实感。"—— 一位娱乐行业从业者
这就是《12星练赛》热度骤降的根本原因。不是制作不够精良，不是技术不够先进——恰恰相反，技术越好、画面越完美，就越暴露了一个核心缺陷：没有灵魂。
AI内容的生存方式，不该是模仿人类
但AI选秀热度下跌，不等于AI内容不行。
问题出在方向错了。很多人做AI内容，第一个想到的就是模仿人类内容——AI唱歌像人、AI跳舞像人、AI演戏像人、AI选秀像真人选秀。好像越像人，就越高级。
这是一个巨大的误区。
AI的优势从来不是"像人"，而是"人做不到的事"。
比如，你可以让AI偶像24小时不间断直播——人类偶像需要睡觉需要休息，AI不需要。这是AI独有的优势。
比如，你可以让观众实时参与AI偶像的人设和剧情走向——人类偶像的性格是固定的，AI偶像的性格可以由观众共同塑造。这也是AI独有的优势。
再比如，低成本、快产出、多语种、多风格——这些都是人类内容做不到或者成本极高的事。
AI选秀如果换一个思路：不是做"AI版的真人选秀"，而是做"只有AI才能做到的选秀"——比如观众可以实时投票改变选手的人设、每周生成全新的挑战主题、甚至观众自己也能生成一个AI角色参与比赛——那它的吸引力可能完全不一样。
目前的AI选秀，本质上是用更先进的技术，在做更低成本的真人选秀替代品。观众新鲜感过了就走，是再正常不过的事。
但AI内容的真正机会，不在于模仿人类内容的样子，而在于探索人类内容不可能做到的形态。
那个方向，才是AI内容的未来。
明天见。
760K likes, down to 50K. In less than two months.
AI talent show "12-Star Training" has 539K followers on Douyin. Its debut stage performance video hit 765K likes, the trailer got 734K - when it first launched, it was huge. AI idols, 12 zodiac personas, talent competition format - every angle hit the algorithm jackpot.
But the heat faded fast. By the August Summer Games series of videos, likes were only 75K and 53K - less than a tenth of the peak.
From 760K to 50K, behind this drop lies a truth many don't want to face: AI can replicate every form, but it can't replicate a soul.
Why Do Talent Shows Need Real People Anyway?
Many people might think, talent shows are just singing and dancing performances, right? AI idols sing better, dance tighter, look more perfect - why aren't audiences buying it?
Because the core appeal of talent shows was never the performances themselves.
What truly hooks people about talent shows is "the sense of growth" - watching an ordinary person transform from nobody to star, seeing every step from awkward to skilled, from nervous to confident, from insecure to self-assured. You vote, you cheer, you stay up late campaigning - essentially, you're participating in someone's growth.
Then there's "the sense of reality" - real people make mistakes, forget lyrics, cry, laugh, freeze up on camera. These "imperfections" are precisely what move people the most. AI idols never make mistakes, they're always perfect - but perfection past a certain point just feels fake.
And most importantly, "charisma and personality" - you like a contestant not usually because they sing the best, but because of their personality, their story, how they talk, how they smile. These things are unique, accumulated over a lifetime of experiences.
AI idols are blank slates. You can give them 12 zodiac archetypes, write every line, design every expression - but these are "designed personalities," not real characters. Audiences find it novel for a few episodes, but after a while they realize: no matter how refined these personas are, fundamentally they're no different from a video game character.
"Do you genuinely cry over a game character's growth? No. After the novelty wears off, you won't genuinely feel for an AI idol's competition either." - An Entertainment Industry Professional
This is the fundamental reason for "12-Star Training's" heat plunge. It's not that production isn't polished enough, not that technology isn't advanced enough - precisely the opposite. The better the tech, the more perfect the visuals, the more it exposes a core flaw: no soul.
AI Content Shouldn't Survive by Imitating Humans
But AI talent show heat dropping doesn't mean AI content doesn't work.
The problem is the wrong direction. A lot of people making AI content immediately think of imitating human content - AI singing like a human, AI dancing like a human, AI acting like a human, AI talent shows like real talent shows. As if the more human-like, the better.
This is a huge misconception.
AI's advantage was never "being like a human" - it's "things humans can't do."
For example, you can have AI idols streaming 24/7 nonstop - human idols need sleep and rest, AI doesn't. That's an AI-exclusive advantage.
For example, you can let viewers participate in real-time in shaping the AI idol's persona and storyline - a human idol's personality is fixed, an AI idol's can be co-created by the audience. Also an AI-exclusive advantage.
And then there's low cost, fast production, multi-language, multi-style - all things human content either can't do or costs a fortune to do.
If AI talent shows took a different approach: not "AI version of a real talent show," but "a talent show only AI could do" - like viewers voting in real time to change contestants' personas, new weekly challenge themes generated on the fly, even viewers generating their own AI characters to compete - the appeal could be completely different.
Current AI talent shows are essentially using more advanced technology to make cheaper substitutes for real talent shows. Audiences leave after the novelty wears off - that's only normal.
But the real opportunity for AI content isn't in imitating what human content looks like - it's in exploring forms human content could never achieve.
That direction is the future of AI content.
See you tomorrow.
AI talent show,12-Star Training,AI idol,virtual idol,AI new media,content creation
Sources · 信源 Sources
本文基于 Dawn Vision 认知引擎处理的 4 个源信号生成，经编辑部人工审核。素材来源：界面新闻/娱乐独角兽。
Generated by the Dawn Vision cognitive engine processing 4 source signals, with human editorial review. Sources: Jiemian News / Entertainment Unicorn.
