<hr /><h1 id="phoenix-reffaellaanima-lora-v2.0-v2.1">菲欧妮 Phoenix Reffaella｜Anima LoRA V2.0 / V2.1</h1>Phoenix Reffaella | Anima LoRA V2.0 / V2.1Two versions are available, each independently trained from Anima Base using the same 55 images, captions, and dataset sampling weights. Choose according to the sample images.- V2.0：sd-scripts：1,500 steps, Rank/Alpha 32/32- V2.1：AI Toolkit：1,500 steps, Rank/Alpha 32/32<h3 id="changes-from-v1"> Changes from V1 </h3>- Expanded the dataset from 40 to 55 images.- Added multi-angle renders of the default outfit.- Added two alternate outfits, 「病名为生」 and 「空想症候」, with dedicated outfit triggers.- Revised captions and dataset sampling weights. Both versions were trained afresh from the base model rather than continuing V1.<h3 id="trigger-words">Trigger words</h3>- Character<code>phoenixreffaella</code>- Default outfit<code>phoenix_default</code>- Alternate outfit — 病名为生<code>phoenix_illness</code>- Alternate outfit — 空想症候<code>phoenix_daydream</code>Combine the character trigger with the desired outfit trigger. The alternate outfits have limited training material, so complex angles and fine clothing details may vary. Additional appearance descriptions can help.The main difference between V2.0 and V2.1 is the trainer and its implementation. V2.1 does not imply proven superiority over V2.0; please refer to the samples when choosing.<h3 id="example-prompts-with-trigger-words">Example prompts with trigger words</h3>These examples expand the training preview prompts with outfit details for V2.0 / V2.1; the revised prompts have not yet been tested. Select one outfit at a time and adjust the pose and background as needed. Circular thigh rings are described separately from the thighhighs.Default outfit portraitphoenixreffaella, phoenix_default, 1girl, solo, blonde hair, yellow eyes, long twintails, hair ornaments, white dress, detached sleeves, portrait, looking at viewer, plain backgroundAlternate outfit: black and redphoenixreffaella, phoenix_illness, 1girl, solo, blonde hair, black and red sailor outfit, white bodysuit, eyepatch, circular thigh rings, white thighhighs, a small gap of exposed upper thighs between the bodysuit and thighhighs, white footwear with red soles, full body, standing, city street, daylightAlternate outfit: blue and whitephoenixreffaella, phoenix_daydream, 1girl, solo, blonde hair, blue and white sailor outfit, cropped white jacket, long white sleeves, navy blue sailor collar, red neckerchief, dark blue bodysuit, white crisscross chest straps, eyepatch, circular thigh rings, white thighhighs, a small gap of exposed upper thighs between the bodysuit and thighhighs, white footwear with red soles and black toe caps, full body, standing, city street, daylightNegative promptlow quality, blurry, bad anatomy, text, watermark<hr />本次提供两个训练版本，均从 Anima Base 独立训练，使用相同的 55 张图片、标签及分组采样比例，可根据展示样图选择。- V2.0：sd-scripts：1500 步，Rank/Alpha 32/32- V2.1：AI Toolkit：1500 步，Rank/Alpha 32/32<h3 id="v1"> 相较 V1 的更新</h3>- 训练素材从 40 张扩充至 55 张。- 补充默认服多角度渲染素材。- 新增「病名为生」和「空想症候」两套换装，并加入独立服装触发词。- 重新整理标注及分组采样比例，从底模新训，非 V1 续训。<h3> 触发词</h3>- 人物<code>phoenixreffaella</code>- 默认服<code>phoenix_default</code>- 病名为生<code>phoenix_illness</code>- 空想症候<code>phoenix_daydream</code>使用时搭配人物与对应服装触发词。两套换装素材较少，复杂角度及服装细节可能出现偏差，可补充具体外观描述。V2.0 与 V2.1 的主要区别是训练器及其实现，V2.1 不代表已经验证全面优于 V2.0，请按样图效果选择。<h3>搭配触发词的提示词示例</h3>以下示例基于训练样图提示词补充了服装细节，可用于 V2.0 / V2.1；修改后的提示词尚未实测。每次选择一套服装，按需替换姿势与背景。圆形腿环使用 circular thigh rings 描述，不作为袜口装饰。默认服肖像phoenixreffaella, phoenix_default, 1girl, solo, blonde hair, yellow eyes, long twintails, hair ornaments, white dress, detached sleeves, portrait, looking at viewer, plain background病名为生：黑红服phoenixreffaella, phoenix_illness, 1girl, solo, blonde hair, black and red sailor outfit, white bodysuit, eyepatch, circular thigh rings, white thighhighs, a small gap of exposed upper thighs between the bodysuit and thighhighs, white footwear with red soles, full body, standing, city street, daylight空想症候：蓝白服phoenixreffaella, phoenix_daydream, 1girl, solo, blonde hair, blue and white sailor outfit, cropped white jacket, long white sleeves, navy blue sailor collar, red neckerchief, dark blue bodysuit, white crisscross chest straps, eyepatch, circular thigh rings, white thighhighs, a small gap of exposed upper thighs between the bodysuit and thighhighs, white footwear with red soles and black toe caps, full body, standing, city street, daylight负面提示词low quality, blurry, bad anatomy, text, watermark<hr />Phoenix Reffaella from Zenless Zone Zero.A character LoRA trained on Anima Base, focusing on her appearance, hairstyle, hair accessories, and default white outfit. Tested in ComfyUI with satisfying results. Feel free to share your creations!Recommended Settings- Base model: Anima Base- Trigger word: phoenixreffaella- Starting strength: 0.8; try adjusting within 0.6–1.0- Resolution: 1024×1024 or a similar pixel count- Sampling steps: 30–40- CFG: 4–5Training Details- Dataset: 40 curated character images and video frames- Training steps: 1,500- Rank / Alpha: 32 / 32- This version focuses on her default outfit. Alternate outfits were excluded from training.Results may vary depending on your base model, prompts, and additional style LoRAs. Try this LoRA on its own first before combining it with others.<hr />示例提示词 / Example Promptmasterpiece, best quality, phoenixreffaella, 1girl, solo,blonde hair, yellow eyes, long twintails, hair ornaments,white dress, black puff sleeves, detached sleeves, red bow,upper body, looking at viewer, gentle smile, outdoors, daylight负向提示词 / Negative Promptworst quality, low quality, blurry, bad anatomy,bad hands, extra fingers, extra limbs, text, watermark---菲欧妮 / Phoenix Reffaella，来自《绝区零》。基于 Anima Base 训练的角色 LoRA，主要学习人物外观、发型、发饰和默认白色服装特征。已在 ComfyUI 中实测，效果满意，欢迎分享返图！使用建议- 底模：Anima Base- 触发词：phoenixreffaella- 推荐起步权重：0.8，可在 0.6～1.0 之间调整- 分辨率：1024×1024 或相近像素量- 采样步数：30～40- CFG：4～5训练信息- 训练素材：40 张经筛选的角色图片及视频截图- 训练步数：1500- Rank / Alpha：32 / 32- 本版本以默认服装为主，其他服装未混入训练。不同底模、提示词和叠加的风格 LoRA 会影响结果，建议先单独加载测试。<hr />角色及原始素材归原权利方所有。The character and original source materials belong to their respective rights holders.

Files with SHA256: 5d16cb3e4aca372a91f4affa1414ebb445ae3d02a024b27101a485b430040976