中文
说实话，如今真正投入时间开发原创素材、建立自己的数据集并训练模型的创作者，可能远比大家想象的要少。社区里有大量模型被冠以“原创训练”或“独立训练”的名义发布，但其中一些实际上只是将其他创作者的模型偷偷融合进自己的权重，再把成果包装成自己的作品。
更令人反感的是，有些人不仅没有尊重原作者，反而以原创开发者的身份宣传自己的模型，借此获取关注、支持，甚至商业利益。如果有人利用未经授权的模型融合来向用户收费，或者向雇主声称自己独立开发了实际上依赖他人成果的模型，那么这已经不只是署名不当的问题，而是诚信与职业操守的问题。
很遗憾，模型盗用已经成为 Civitai 上一个相当严重的问题。许多热门模型在未经我许可的情况下，将一个或多个 Moody 系列模型直接融入了自身权重之中。更令人担忧的是，部分创作者甚至普通用户可能根本不知道这件事的存在。因为他们使用的模型，可能本身就是由其他未经授权、已经合并了我的模型的作品再次合并而成的。就这样，我的原创工作可能在不知情的情况下，被一层又一层地传播出去。
我认为，真正的原创模型开发者值得尊重，因为他们投入了时间、精力和资源去创造新的东西。如果一个社区逐渐变成了谁更擅长拿走别人的成果、隐藏模型来源、再把它包装成自己的作品，谁就能获得更多利益，那么最终受到伤害的，将是所有认真创作的人。
我想特别说明的是：我并不反对别人使用或融合我的模型，我反对的是未经许可就擅自使用，并把别人的劳动成果据为己有。
事实上，如果有人愿意提前联系我、坦诚说明自己的用途并征求授权，我一直都很愿意沟通。到目前为止，我还没有拒绝过任何向我提出授权请求的人。
所以，问题从来不是我不愿意分享，也不是我想阻止别人进步。真正让我无法接受的，是某些人理所当然地认为自己有权拿走别人的成果，既不征求许可，也不提供应有的署名，甚至进一步将这些成果商业化，借此为自己谋利。
愿意分享，不代表可以被随意利用；愿意授权，也不代表别人可以跳过授权这一步。
我尊重真正的原创者，也尊重愿意坦诚沟通、遵守规则的开发者。但对于那些明明依赖他人的成果，却刻意隐瞒来源、冒充原创，甚至借此牟利的人，我认为社区完全有理由提出质疑，并要求他们承担应有的责任。
English
Honestly, I believe there are far fewer genuine original content creators in this community than most people realize. Very few invest the time and effort to develop original source material, build their own datasets, and train models from the ground up. Yet many models are presented as independently trained or original creations, while some are actually built by quietly merging other creators' work into their weights and then claiming the results as their own.
Most current Krea2 popular models contain one or more Moody models (theft events are not limited to moody models) merged into their weights without my consent. Even worse, some creators may unknowingly use these models, not realizing that my work has been passed down through multiple unauthorized merges.
What makes this particularly frustrating is when people use that false image of originality to gain recognition, attract supporters, or make money. If someone uses unauthorized model merges to charge users or tells an employer they independently developed a model that actually relies on other creators' work, that goes beyond improper attribution. It raises serious questions about honesty and professional integrity.
Genuine model developers deserve recognition for the time, effort, and resources they invest in creating something of their own. If this community becomes a place where people can profit by taking others' work, concealing its origins, and repackaging it as their own, the people who ultimately suffer are the creators who genuinely contribute to the community.
I want to make one thing absolutely clear: I have no fundamental problem with people using or merging my models. My problem is with people taking my work without permission and claiming it as their own.
In fact, if someone approaches me respectfully, explains what they want to do, and asks for permission beforehand, I am generally more than happy to discuss it. So far, I have never turned down anyone who has actually approached me to request permission.
This has never been about refusing to share or trying to prevent others from improving their models. What I cannot accept is the entitlement of people who assume they have the right to take someone else's work without asking, provide no proper attribution, and sometimes go even further by commercializing it for their own benefit.
Being willing to share does not mean being free to exploit. Being willing to grant permission does not mean permission can simply be ignored.
I respect genuine creators, and I respect developers who communicate honestly and follow the rules. But when people knowingly rely on someone else's work while concealing its origins, falsely presenting it as their own, or profiting from it without authorization, I believe the community has every right to question their conduct and hold them accountable.
I remember someone in a group once pointing out a strange phenomenon with open models: after a model gets released, especially around the middle of its lifecycle, it somehow starts feeling more and more “AI-like.” But when you go back and try the original base model, you realize it actually feels completely different from all the newer versions.
I think a big part of the reason is that everyone keeps merging and mixing models with each other. In fact, we've already discovered cases where models from different authors have been indirectly merged with one another, without the authors even realizing it.
Eventually, the entire model ecosystem turns into one giant melting pot of recycled influences. And when you keep mixing every color together, the final result is inevitably going to be the color of 💩.
我记得以前群里有人吐槽过一个挺神奇的现象：一个开放模型发布之后，尤其是到了生命周期的中期，你会发现它的各种衍生版本越来越有那种浓浓的“AI味”。但当你回过头去用最初发布的原版模型时，却会惊讶地发现，原版和现在这些新模型的感觉完全不一样。
我觉得很大一部分原因，就是大家一直在互相融合、互相借鉴，模型之间不断交叉污染。事实上，我们已经发现了一些案例：不同作者的模型在他们自己都不知情的情况下，就已经通过各种间接的方式互相融合了。
到最后，整个模型生态就变成了一个巨大的染缸，所有颜色都被倒进去搅在一起。
而当你把所有颜色都混在一起的时候，最后出来的颜色，大概率就是💩的颜色吧。