薇莉丝，漫画《满级穿越的我为什么是牧师小姐啊》的主角。在部分翻译中也被称为Willis。Violet, the protagonist of the manga I Am a Max-level Priestess in Another World. Also known as Willis in some translations.<hr />Trigger Word (触发词):<pre><code>Violet Priestess</code></pre><pre><code>breasts, long hair, detached sleeves, ahoge, multicolored hair, small breasts, hairband, clothing cutout, dress, headband, bare shoulders, white hairband, white dress, center opening, black hair, two-tone hair, white thighhighs, skirt, small cross-shaped ornaments on clothes, cross-shaped pupils, openwork design sun cross on the collar</code></pre><pre><code>underwear, striped panties, high heel boots</code></pre>请在负面提示词（Negative Prompt）中加入：<code>coattails</code>Please add to the negative prompts: <code>coattails</code>更多详情请参考示例图片中的ComfyUI工作流。For more details, please refer to the ComfyUI workflow in the example images.<hr />Anima 模型很有意思的一点在于它对自然语言的处理能力。你可能已经注意到，提示词里有些部分和传统的 Danbooru 标签不太一样，反而更接近日常短语。但实际上，我在训练时依然采用了类似 IllustriousXL 的常规打标（Tags）方法。我把这称为“训练后的提示词工程”。得益于大语言模型（LLM）的识图能力（如果你不知道如何用短语描述，可以使用 Gemini 等工具，把图片发给它并告诉它你想实现的效果）以及 Anima 自身对短语的理解力，你可以通过调整提示词来更完美地还原角色特征，非常推荐你也亲自尝试一下！One interesting aspect of the Anima model is its natural language processing capability. As you can see in the prompts, some parts differ from traditional Danbooru tags and resemble short phrases instead. However, during training, I actually used the standard tagging method, similar to IllustriousXL. I call this "post-training prompt engineering." By leveraging LLM image captioning (if you don't know how to describe something with phrases, you can use tools like Gemini by providing an image and telling it what you want) alongside Anima's phrase understanding, you can fine-tune your prompts to better reproduce specific features. I highly recommend giving it a try!<hr />Important reminder: Do not use for commercial purposes or publish on other public platforms重要提示：请勿商用并发表到其他公开平台

Files with SHA256: ccc734a948f6c40b8516e4993334e7978d1466e17d36bed001870e2de40c0acd