新的藿藿lora将重点放在了泛化性上，同时有着非常不错的还原能力，我会将我的lora模型提示词的使用方法放在下方，首先要说明的是一些基础问题：1.训练底模使用了NAI，lora具有更自由的发挥空间，你可以使用真人类、二次元以及各种风格的底模生成不同的藿藿，二次元推荐meinamix系列、anyloraCheckpoint、cetus、counterfit系列。当然不局限于这些，这些模型是我经常使用的。<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/7d4b1d64-5f3d-40ea-a632-f685bffc443f/width=525/7d4b1d64-5f3d-40ea-a632-f685bffc443f.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/02b31eff-7f47-49f7-8935-b449e5d8c333/width=525/02b31eff-7f47-49f7-8935-b449e5d8c333.jpeg" />2.huohuoV5可以搭配其它的lora使用,并且有不错的效果，比如Q版手办、像素风、表情包、特定服装、动作等。<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fe978e18-01df-46b4-a722-a7b0c274ef67/width=525/fe978e18-01df-46b4-a722-a7b0c274ef67.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f5943a7c-9317-4b3a-bf6d-748631d8cdf6/width=525/f5943a7c-9317-4b3a-bf6d-748631d8cdf6.jpeg" />3.所有的部分你都可以自由改变（除了藿藿的发色发型以及眼睛，这些是我为藿藿保留的最基本特征），比如生成各种背景、各种视角、各种动作的图片。<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/43ceba78-0833-4dce-9581-f86b49c30862/width=525/43ceba78-0833-4dce-9581-f86b49c30862.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/2e2108c3-4980-4138-a844-aabdf7d6eb4c/width=525/2e2108c3-4980-4138-a844-aabdf7d6eb4c.jpeg" />4.以下是关于tag的使用说明：前置：huohuo,1girl,solo藿藿的帽子：jiangshi_hat发饰：yinyang_hair ornament呆毛：ahoge藿藿的耳朵：animal ears胸前的太极玉佩：tassel jade藿藿的衣服：全部融入进了open dress，long sleeves,detached sleeves,你可以删除这些词使用其他词语完成换装藿藿的短裤：shorts藿藿的堆堆袜：loose socks藿藿的鞋子：shoes藿藿的尾巴：weiba_fire tail一些符咒：spell面部表情：自由度很高，你甚至可以使用&gt; &lt;背景：自由发挥，想生成白背景使用simple background,white backgroundThe new Huohuo Lora focuses on generalization and has excellent restoration ability. I will place the usage of prompt words in my Lora model below, first of all, to clarify some basic issues:1. The training bottom mold uses NAI, and lora has more freedom to play. You can use real human, anime and various styles of bottom molds to generate different hops. anime recommends meinamix series, anyloraCheckpoint, cetus, counterfit series. Of course, it's not limited to these, these models are what I often use.2. HuohuoV5 can be used with other Loras and has good effects, such as Q-version handmade, pixel style, emoticons, specific clothing, actions, etc.3. You can freely change all parts (except for the hair color, hairstyle, and eyes of Huo Huo, which are the most basic features I have reserved for him), such as generating images of various backgrounds, perspectives, and actions.4. The following are instructions for using tags:Front: huohuo, 1Girl, soloHuohuo's hat: Jiangshi_ HatHairdressing: yinyang_ Hair omentAogeAnimal ears of HuohuoTai Chi Jade pendant on the chest: tassel jadeHuohuo's clothes: all integrated into open dress, long sleeps, detached sleeps, you can delete these words and use other words to complete the dressing changeHuohuo's shortsLoose socksShoesThe tail of Huohuo: weiba_ Fire tailSome spells: spellFacial expression: High degree of freedom, you can even use&gt;&lt;Background: Free play, want to generate a white background, use simple background, white background

Honkai:Star Rail-Huohuo(崩坏：星穹铁道 藿藿)

huohuov5.safetensors

huohuov5

The following are instructions for using tags:Front: huohuo, 1Girl, soloHuohuo's hat: Jiangshi_ HatHairdressing: yinyang_ Hair oment, ahoge, animal earsTai Chi Jade pendant on the chest: tassel jadeHuohuo's clothes: all integrated into open dress, long sleeps, detached sleeps, shorts, loose socks, shoesyou can delete these words and use other words to complete the dressing changeThe tail of Huohuo: weiba_fire tail or weiba fire tailSome spells: spellFacial expression: High degree of freedom, you can even use&gt;&lt;Background: Free play, want to generate a white background, use simple background, white backgroundThe new Huohuo Lora focuses on generalization and has excellent restoration ability. I will place the usage of prompt words in my Lora model below, first of all, to clarify some basic issues:1. The training bottom mold uses NAI, and lora has more freedom to play. You can use real human, anime and various styles of bottom molds to generate different hops. anime recommends meinamix series, anyloraCheckpoint, cetus, counterfit series. Of course, it's not limited to these, these models are what I often use.2. HuohuoV5 can be used with other Loras and has good effects, such as Q-version handmade, pixel style, emoticons, specific clothing, actions, etc.3. You can freely change all parts (except for the hair color, hairstyle, and eyes of Huo Huo, which are the most basic features I have reserved for him), such as generating images of various backgrounds, perspectives, and actions.

Files with SHA256: 32db1c5f37bd50935f27d4ef3e339d3acf26281984b1c1d8cf76df8f559a0c6c