东方韵味:Stable Diffusion极致中国风LoRA模型优选推荐

模型特点:中国敦煌飞天仙女

图片

Prompts: 8K raw photo, Best quality, masterpiece, ultra high res, (zentangle, mandala, tangle, entangle),colorful,(photorealistic:1.4), raw photo, 1girl, offshoulder, outdoors,tall female, skinny, slim_legs, solo,(perfect body shape),holy light,gold foil,gold leaf art,glitter drawing,fly high in the sky ,(full body),long hair, (see-through:1.3), ((revealing clothes)), ((barefoot)), medium breasts, earrings, ((dress)), no bra, ((bare shoulders)), very long hair, no panties, black hair, ((detached sleeves)), ((side slit)),bare,dunhuang_cloths,dunhuang_dress,traditional chinese hair style,Traditional Chinese hair ornament,cinema,Fei Tian Ji ,Propitious Clouds,chang'e, paradox,pure,soft smile,illustration,barefoot,(pelvic curtain),playful patterns, lively texture

Checkpoint:darkSushiMixMix_brighterPruned

图片

Prompts: (masterpiece, top quality, best quality, official art, beautiful and aesthetic:1.2), (1girl:1.3), (fractal art:1.3), extreme detailed, colorful,highest detailed,illustrations, artistic fusion,fantastical scenes,striking visuals, dunhuang_style, dunhuang_cloths,dancing, desert, long floating hair, jewelry,floating ribbon,wind,sunset,

Checkpoint:ghostmix_v20Bakedvae

图片

Prompts:(RAW photo, best quality),(realistic, photo-realistic:1.3),best quality,highly detailed,masterpiece,ultra-detailed,illustration,1girl,upper_body,dynamic angle,world masterpiece theater,messy_long_hair,best quality,extremely detailed CG unity 8k wallpaper,ink,Amazing,cinematic lighting,lens_flare,,dunhuang_style

Checkpoint:darkSushiMixMix_brighterPruned

模型介绍:

作者目前训练了1.0和2.0两个版本,2.0训练级改为512x768,更适合做竖图;训练方式改为locon 128dim,增加了更多服饰颜色。目前问题是手部依然不稳定,建议可以尝试下 controlnet 或加 embedding 修手。

模型说明:

- Trigger Words:

DUNHUANG_CLOTHS

DUNHUANG_STYLE

DUNHUANG_BACKGROUND

DUNHUANG_DRESS

DUNHUANG_FAN

- Checkpoint:动漫真人的模型都可以,人像模型兼容性都挺强的

MoXin

模型特点:水墨画,写意,工笔

图片

Prompts: shukezouma, negative space, , shuimobysim ,, portrait of a woman standing , willow branches, (masterpiece, best quality:1.2), traditional chinese ink painting,, modelshoot style, peaceful, (smile), looking at viewer, wearing long hanfu, hanfu, song, willow tree in background, wuchangshuo

Checkpoint: chilloutmix_NiPrunedFp32Fix

图片

Prompts: shukezouma, negative space, , shuimobysim ,,a branch of flower, traditional chinese ink painting

Checkpoint:chilloutmix_NiPrunedFp32Fix

图片

Prompts: masterpiece, best quality, 1girl,black hair,red eyes, hanfu,scabbard,wuxia,sword, blood, bamboo forest,enemy

Checkpoint:ghostmix_v12

图片

Promptsblack hair,red eyes, hanfu,scabbard,wuxia,sword, bamboo forest,enemy, negative space, shuimobysim , portrait of a woman standing , (masterpiece, best quality:1.2), traditional chinese ink painting , peaceful, looking at viewer, wearing long hanfu, hanfu

Checkpoint:OM3A1B_orangemixs

模型介绍:

墨心整个系列模型有三个版本,分别是疏可走马1.0、疏可走马1.1和墨心1.0,作者介绍墨心1.0主要由安吉吴仓石、兴化板桥先生、八大山人、山阴伯年等大师作品辅以现代人物训练而成;疏可走马是一个和墨心搭配使用的构图LoRA, 一旦使用并再最前前置提示词后,就会采用较大面积留白的构图风格。

模型说明:

作者推荐使用参数

- CFG 1~3 : 大小写意

- CFG 3~7 : 逐渐工笔

- 墨心 LoRA Weight:<0.85

- 疏可走马 LoRA Weigh:0.7~1

- Checkpoint 推荐:ChilloutMix、Guofeng3.2等

Xiaorenshu

模型特点:中国连环画风格,重线条,风景写意,人物沁彩

图片

Prompts:(illustration:1.1),(best quality),(masterpiece:1.1),1gilr,standing,flowers,bush,outdoors,looking at viewer

图片

Prompts:(upper body: 1.5),(white background:1.4), ((tachi-e)), original,(illustration:1.1),(best quality),(masterpiece:1.1),(extremely detailed CG unity 8k wallpaper:1.1), (colorful:0.9),(panorama shot:1.4),(full body:1.05),(solo:1.2), (ink splashing),(color splashing),((watercolor)), clear sharp focus,{ 1girl standing },((chinese style )),(flowers,woods),outdoors,rocks, looking at viewer, make happy expressions,soft smile,pure,beautyfull detailed face and eyes,beautyfull intricacy clothing decorative pattern details, black hair,black eyes,colorwater

图片

Prompts: ((best quality)), ((masterpiece)), (detailed), realistic, multiple boys,(riding motorcycle),mustache,chinese clothes,armour,chinese general,looking at viewer, full body,night,moon in the night sky,outdoors,motorcycle

图片

Prompts: (illustration:1.1),(best quality),(masterpiece:1.1), shz,1man,Wu_Song,headband,(ridding on a solo tiger+punching the head),solo,forest,full moon, detailed face

模型介绍:

该 LoRA 风格正如他的命名:连环画,作者介绍这是根据 20 世纪初叶在上海形成并开始广泛流传的通俗图画读物连环画炼的,图片效果具有很强的故事性,人物及风景线条浓厚,颜色偏向于灰黄调。该 LoRA 辨识度非常强,除了一些古典元素的 prompt 能识别,近现代的元素比如摩托车、蜘蛛侠之类的识别也很准确。

模型说明:

v2.0版本:同1.5版本

- v1.5版本:CFG<8,权重<=0.9

- v1.0版本:CFG=3,权重=0.9;一些二次元模型使用高CFG后会得到一种奇妙的抽色效果

The Three Kingdoms

模型特点:光荣游戏《三国志》人物形象

图片

Prompts: (4k, best quality, highres:1.1), (masterpiece:1.1), man, (Chinese male:1.2), middle-aged, warrior, detailed eyes, facial hair, muscular, strong, brave, courageous, powerful, honorable, disciplined, confident, looking at viewer, face, bracer, helmet, armor, holding weapon, riding, horse, black, red

Checkpoint:CounterfeitV25_25

图片

Prompts: (4k, best quality, highres:1.1), (masterpiece:1.1), man, Chinese male, detailed eyes, looking at viewer, (Collin Chou:1.5), male_focus, , 1boy, facial_hair, brown_eyes, katana, chinese_clothes, crew cut, muscular, wide shoulder, topknot, confident

Checkpoint: CounterfeitV25_25

图片

Prompts: (4k, best quality, highres:1.1), (masterpiece:1.1), man, Chinese male, detailed eyes, looking at viewer, (Chen Daoming:1.6), facial_hair, hanfu, 1boy, beard, male_focus, mustache, chinese_clothes, old, solo, old_man, wide_sleeves, own_hands_together, long_sleeves, topknot, robe, looking_at_viewer, wrinkled_skin

Checkpoint: CounterfeitV25_25

模型介绍:

如题,是关于《三国志》游戏人物角色立绘,「大江东去浪淘尽,千古风流人物」。作者目前已经更新了四个版本,下面是部分版本的更新介绍。

Version1.1更新:

- 部分解决文官女性化问题,仍需要添加负面

- tag:earrings, hair flower, female clothes, female ornament, female

一个简单的脸谱化方法:直接把人名加prompt里,参见样例。

Version1.0更新:

- 新加入了三国志战略版武将

- 使用新的底模:CounterfeitV25_25.safetensors 注意要点选V2.5,作者现在更新到3.0了

- 由于新底模机制不同,不要添加类似 realistic, photo-realistic 的 prompt

- 素材分辨率提升至最高768x1024

- 删去了阴险脸(如李儒)和异民族,新删去黄巾贼

- 样例使用Eular a + Latent(Nearest-exact),Denoising Strength=0.5

- 看板娘照例使用了其他Lora和Controlnet调脸,理论上男武将也可以这样操作

- 已知问题:由于女性角色数量激增,男性文官生成中可能出现女性化,暂时添加负面prompt来workaround:earrings, hair flower, female clothes, female ornament

Version0.999更新:

- 新加入了近200名武将

- 删去了阴险脸(如李儒)和异民族服饰

- Denoising Strength=0.5

模型说明:

- Checkpoint:Ver.0.999推荐guofeng3_v32或其Light版本;Ver.1推荐CounterfeitV25

- Size: Ver.0.999推荐512 x 768;Ver.1推荐768x1024

- Weight:  1

- Sampler:Eular a

- Prompts:加上 chinese male, chinese female, chinese

Chinese Traditional 

Clothing 

特点:中国传统服饰合集

图片

Prompts:(long skirt:1.2),(blue flowers print:1.2),(yellow flowers print:1.2),(red flowers print:1.3),(birds print:1.2),(leaf print:1.2),1 girl,full body, (short hair:1.1), (realistic:1.7),((best quality)),absurdres,(ultra high res),(photorealistic:1.6),photorealistic,octane render,(hyperrealistic:1.2), (photorealistic face:1.2), (8k), (4k), (Masterpiece),(realistic skin texture), (illustration, cinematic lighting,wallpaper),( beautiful eyes:1.2),((((perfect face)))),(cute),(standing),(black hair),(short hair),black eyes,red lips, (outdoors)

Checkpoint:majicmixRealistic_v4

图片

Prompts:(blue/red/green/white/yellow/purple printed chinese upshirt),(long skirt),1 girl,upper body, high heels,(short hair:1.1), (realistic:1.7),((best quality)),absurdres,(ultra high res),(photorealistic:1.6),photorealistic,octane render,(hyperrealistic:1.2), (photorealistic face:1.2), (8k), (4k), (Masterpiece),(realistic skin texture), (illustration, cinematic lighting,wallpaper),( beautiful eyes:1.2),((((perfect face)))),(cute),(standing),(black hair),(short hair),black eyes,red lips, (outdoors)

Checkpoint:majicmixRealistic_v4

图片

Prompts: (pink/yellow/green/blue/red/white chinese clothes),(floral print), (pink/yellow/green/blue/red/white/black long skirt),(long sleeves),1 girl,full body,(short hair:1.1), (realistic:1.7),((best quality)),absurdres,(ultra high res),(photorealistic:1.6),photorealistic,octane render,(hyperrealistic:1.2), (photorealistic face:1.2), (8k), (4k), (Masterpiece),(realistic skin texture), (illustration, cinematic lighting,wallpaper),( beautiful eyes:1.2),((((perfect face)))),(cute),(standing),(black hair),black eyes,red lips, indoors, chinese ornaments, wooden chinese table, window, flowers

Checkpoint:majicmixRealistic_v4

图片

Prompts:(hanfu),(floral print:1.3),(pink see-through shirt),(green see-through print long skirt),(long sleeves),1 girl,full body,(long hair:1.1), (realistic:1.7),((best quality)),absurdres,(ultra high res),(photorealistic:1.6),photorealistic,octane render,(hyperrealistic:1.2), (photorealistic face:1.2), (8k), (4k), (Masterpiece),(realistic skin texture), (illustration, cinematic lighting,wallpaper),( beautiful eyes:1.2),((((perfect face)))),(cute),(standing),(black hair),black eyes,red lips, outdoors, (chinese style buildings)

Checkpoint:majicmixRealistic_v4

图片

Prompts:(hanfu),(chinese clothes),((floral print)),((yellow magnificent print long sleeves)),((green print see-through dress)),1 girl,full body,(short hair),(realistic:1.7),((best quality)),absurdres,(ultra high res),(photorealistic:1.6),photorealistic,octane render,(hyperrealistic:1.2), (photorealistic face:1.2), (8k), (4k), (Masterpiece),(realistic skin texture), (illustration, cinematic lighting,wallpaper),( beautiful eyes:1.2),((((perfect face)))),(cute),(standing),(black hair),black eyes,red lips, outdoors,(chinese style countryard),(pond)

Checkpoint:majicmixRealistic_v4

图片

Prompts:(hanfu),(chinese clothes),((floral print)),((pink/yellow/blue/white/red/black print chinese shirt)),((pink/yellow/blue/white/red/black print long skirt)),1 girl,full body,(short hair),(realistic:1.7),((best quality)),absurdres,(ultra high res),(photorealistic:1.6),photorealistic,octane render,(hyperrealistic:1.2), (photorealistic face:1.2), (8k), (4k), (Masterpiece),(realistic skin texture), (illustration, cinematic lighting,wallpaper),( beautiful eyes:1.2),((((perfect face)))),(cute),(standing),(black hair),black eyes,red lips, outdoors,(chinese style countryard)

Checkpoint:majicmixRealistic_v4

图片

Prompts:1girl,floral print, (yellow print chinese shirt),(green print long skirt),full body, short hair, outdoors, park, lawn, flowers,trees, sunshine,lake

Checkpoint:majicmixRealistic_v4

模型介绍:

这是一个中国传统服饰合集的 LoRA,特点在于集合了不同时期的传统服饰:马面裙、旗袍以及魏晋、唐、宋服装,在此基础上又进行了一些删改和创新。这些 LoRA 配合真人大模型效果更佳,比如 majicmixrealistic 之类的,动漫水彩风有时候也会有出人意料的效果。

模型说明:

Checkpoint推荐:majicmixRealistic_v4

Trigger words:

Horse-Face skirt: light green/red/blue/yellow/purple long skirt,birds/flowers/leaf print

CheongsamStyle1:china dress, dress, green printed china dress, purple plaid china dress and blue printed plaid china dress

CheongsamStyle2:blue/yellow/red/green/white chinese upshirt, floral/birds printed chinese upshirt, blue/yellow/green/white long skirt

QingNvZhuangV1.0:pink/yellow/green/blue/red/white chinese clothes,pink/yellow/green/blue/red/white/black long skirt, floral print, long sleeves

Hanfu_Song:hanfu, floral print, long shirt, long skirt, long sleeves, long green shirt , long white skirt

Hanfu_Tang:(hanfu),(chinese clothes),((floral print)),((green/blue/yellow/pink/white/blackmagnificent print long sleeves)),((pink/white/black/green/blue/yellow print see-through dress))

Hanfu_Wei Dynasty: (hanfu),(chinese clothes),((floral print)),((pink/yellow/blue/white/red/black print chinese shirt)),((pink/yellow/blue/white/red/black print long skirt))

The new version of Hanfu_weijin:1girl,floral print, (blue/white/red/black/yellow/pink print chinese shirt),(pink/yellow/blue/white/red/black print long skirt)

Weight: 推荐0.6~0.7

如何训练LorA

对于很多刚学习AI绘画的小伙伴而言,想要提升、学习新技能,往往是自己摸索成长,不成体系的学习效果低效漫长且无助。

如果你苦于没有一份Lora模型训练学习系统完整的学习资料,这份网易的《Stable Diffusion LoRA模型训练指南》电子书,尽管拿去好了。

包知识脉络 + 诸多细节。节省大家在网上搜索资料的时间来学习,也可以分享给身边好友一起学习。

由于内容过多,下面以截图展示目录及部分内容,完整文档领取方式点击下方微信卡片,即可免费获取!

 

img

img

img

篇幅有限,这里就不一一展示了,有需要的朋友可以点击下方的卡片进行领取!

 

  • 15
    点赞
  • 14
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
要训练自己格的模型,可以使用lora-scripts目录中的train.ps1脚本文件。在脚本中,你需要设置以下参数来进行训练: - $pretrained_model:这是你下载的训练模型的基础模型,你可以将其替换为你想要使用的模型。\[1\] - $train_data_dir:这是你的训练数据目录,你可以将其设置为包含你的训练数据的文件夹路径。\[1\] - $resolution:这是图像的分辨率,你可以设置为你想要的宽度和高度。注意,宽度和高度必须是64的倍数,并且可以是非正方形的。\[3\] - $batch_size:这是每个训练批次的样本数量,你可以根据你的需求进行设置。\[3\] - $max_train_epoches:这是最大的训练轮数,你可以根据需要进行设置。\[3\] - $save_every_n_epochs:这是每隔多少个训练轮数保存一次模型。你可以根据需要进行设置。\[3\] 通过调整这些参数,你可以训练出符合你自己格的模型。请确保你的训练数据和模型选择与你想要的格相匹配。\[2\] #### 引用[.reference_title] - *1* *3* [stable diffusion打造自己专属的LORA模型](https://blog.csdn.net/wangqingpei557/article/details/130049592)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v91^control,239^v3^insert_chatgpt"}} ] [.reference_item] - *2* [Stable Diffusion Lora模型训练详细教程](https://blog.csdn.net/weixin_45250844/article/details/130302817)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v91^control,239^v3^insert_chatgpt"}} ] [.reference_item] [ .reference_list ]

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值