要点初见:Stable Diffusion NovelAI模型优质文字Tag汇总与实践魔咒汇总
Posted BingLiHanShuang
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了要点初见:Stable Diffusion NovelAI模型优质文字Tag汇总与实践魔咒汇总相关的知识,希望对你有一定的参考价值。
目前贴吧、B站上有大量Stable Diffusion的模型资源、TAG、TAG生成器分享,其中居然有不少试图靠信息差把这些开源资源卖钱的。加上目前网上相关的TAG整理贴极少,不少TAG也是以图片的形式存在,故博主打算在此专门整理一下网友们实践过的成套的“魔咒”,并在自己的1060 6G上出图验一验效果。(不同的推理显卡可能会有不同的输出)
本文聚焦于基于NovelAI模型的二次元图像生成,不涉及原版stable-diffusion-v1-4模型(目前v1-5模型站在Stable Diffusion的两家创始公司StabilityAI与Runway纷争的风口浪尖上hhh),不涉及TAG生成器。
一、前情提要
Stable Diffusion框架:stable diffusion webui版本
框架Github链接:https://github.com/AUTOMATIC1111/stable-diffusion-webui
模型checkpoint(从下述模型部署方案中下载):final-pruned.ckpt [925997e9],3.97 GB (4,265,377,440 字节)
本文采用的模型部署方案:【AI绘画】超详细novelai教程!解压即用,超分无损图片以及各种风格关键词!_哔哩哔哩_bilibili
关于Stable Diffusion webui版本输入词组的加权减权:
在Stable Diffusion webui版本中使用()对词语权重进行增加,使用[]对词语权重进行减少,从源码中看没有加权重的效果,代码位置位于modules/prompt_parser.py,Readme里亦有提及此attention机制。()能在原权重上乘1.1,[]则是将原权重除以1.1:
def parse_prompt_attention(text):
"""
Parses a string with attention tokens and returns a list of pairs: text and its associated weight.
Accepted tokens are:
(abc) - increases attention to abc by a multiplier of 1.1
(abc:3.12) - increases attention to abc by a multiplier of 3.12
[abc] - decreases attention to abc by a multiplier of 1.1
\\( - literal character '('
\\[ - literal character '['
\\) - literal character ')'
\\] - literal character ']'
\\\\ - literal character '\\'
anything else - just text
>>> parse_prompt_attention('normal text')
[['normal text', 1.0]]
>>> parse_prompt_attention('an (important) word')
[['an ', 1.0], ['important', 1.1], [' word', 1.0]]
>>> parse_prompt_attention('(unbalanced')
[['unbalanced', 1.1]]
>>> parse_prompt_attention('\\(literal\\]')
[['(literal]', 1.0]]
>>> parse_prompt_attention('(unnecessary)(parens)')
[['unnecessaryparens', 1.1]]
>>> parse_prompt_attention('a (((house:1.3)) [on] a (hill:0.5), sun, (((sky))).')
[['a ', 1.0],
['house', 1.5730000000000004],
[' ', 1.1],
['on', 1.0],
[' a ', 1.1],
['hill', 0.55],
[', sun, ', 1.1],
['sky', 1.4641000000000006],
['.', 1.1]]
"""
res = []
round_brackets = []
square_brackets = []
round_bracket_multiplier = 1.1
square_bracket_multiplier = 1 / 1.1
def multiply_range(start_position, multiplier):
for p in range(start_position, len(res)):
res[p][1] *= multiplier
for m in re_attention.finditer(text):
text = m.group(0)
weight = m.group(1)
if text.startswith('\\\\'):
res.append([text[1:], 1.0])
elif text == '(':
round_brackets.append(len(res))
elif text == '[':
square_brackets.append(len(res))
elif weight is not None and len(round_brackets) > 0:
multiply_range(round_brackets.pop(), float(weight))
elif text == ')' and len(round_brackets) > 0:
multiply_range(round_brackets.pop(), round_bracket_multiplier)
elif text == ']' and len(square_brackets) > 0:
multiply_range(square_brackets.pop(), square_bracket_multiplier)
else:
res.append([text, 1.0])
for pos in round_brackets:
multiply_range(pos, round_bracket_multiplier)
for pos in square_brackets:
multiply_range(pos, square_bracket_multiplier)
if len(res) == 0:
res = [["", 1.0]]
# merge runs of identical weights
i = 0
while i + 1 < len(res):
if res[i][1] == res[i + 1][1]:
res[i][0] += res[i + 1][0]
res.pop(i + 1)
else:
i += 1
return res
关于隐形水印和nsfw:
webui版在代码中把隐形水印开启默认设置为False,而nsfw需要自行添加在负面TAG中才会起到避免生成不正经图的功能。
二、文字TAG汇总
所有TAG都经过格式核对,但可能有重复语句。本文中的图像都是模型输出的原图,若需高清化可使用超分算法。
1、爱丽丝
来自AmberK♬的贴吧的分享,一位专门研究用NovelAI模型画爱丽丝的贴吧用户,出图效果非常惊艳,形意景致完美配合,堪称教学案例。附上这位老哥的出图心得:
【图片】【教学】召唤学派-魔咒基础讲义(NovelAI Tag词条基础)【novelai吧】_百度贴吧
TAG:
Positive prompt:
((alice)), alice in wonderland, (((solo))), 1girl, ((delicate face)), vely long hair, blunt_bangs, (((full body))), (floating hair), (looking_at_viewer), open mouth, (looking_at_viewer), open mouth, blue eyes, Blonde_hair, Beautiful eyes, gradient hair, ((white_frilled_dress)), ((white pantyhose)), (long sleeves), (juliet_sleeves), (puffy sleeves), white hair bow, Skirt pleats, blue dress bow, blue_large_bow, (((stading))), (((arms behind back))), sleeves past wrists, sleeves past fingers, (forest), flowering hedge, scenery, Flowery meadow, clear sky, (delicate grassland), (blooming white roses), flying butterfly, shadow, beautiful sky, cumulonimbus, ((absurdres)), incredibly_absurdres, huge_filesize, (best quality), (masterpiece), delicate details, refined rendering, original, official_art, 10s,
Negative prompt:
lowres, highres, worst quality, low quality, normal quality, artbook, game_cg, duplicate, grossproportions, deformed, out of frame, 60s, 70s, 80s, 90s, 00s, ugly, morbid, mutation, death, kaijuu, mutation, no hunmans, monster girl, arthropod girl, arthropod limbs, tentacles, blood, size difference, sketch, blurry, blurry face, blurry background, blurry foreground, disfigured, extra, extra_arms, extra_ears, extra_breasts, extra_legs, extra_penises, extra_mouth, multiple_arms, multiple_legs, mutilated, tranny, trans, trannsexual, out of frame, poorly drawnhands, extra fingers, mutated hands, poorly drawn face, bad anatomy, bad proportions, extralimbs, more than 2 nipples, extra limbs, bad anatomy, malformed limbs, missing arms, miss finglegs, mutated hands, fused fingers, too many fingers, long neck, bad finglegs, cropped, bad feet, bad anatomy disfigured, malformed mutated, missing limb, malformed hands,
其它参数:
Steps: 50
Sampler: DDIM
CFG scale: 7
Size: 1024 x 1024
博主实践与心得:
个人最喜欢这款TAG的画风,爱丽丝的人物特征、动作、背景和表情均无可挑剔,除了亮度略高需要后期简单降一下之外,图片生成稳定性超出我的想象:批量出图中仅偶尔会出现腿部缺失、手部细节错误、手臂缺失的问题(NovelAI的老毛病了),且这些需要修改的细节占比非常低,不愧是爱丽丝爱好者专门调出来的好TAG!
非常推荐!
2、星空下邪笑的马猴烧酒
来自HBV非的贴吧的分享,学习的是hiten老师的画风,吧友们试了都说好!
原贴链接:
TAG:
Positive prompt:
hiten_1, ((masterpiece)), (((best quality))), ((ultra-detailed)), ((illustration)), ((disheveled hair)), beautiful detailed eyes, (1girl:1.2),(solo), dynamic angle, dark magician girl,(black kneehighs:1.2),(starry tornado:1.4), (starry Nebula:1.1), star, ((frills)), beautiful detailed sky, beautiful detailed eyes,evil smile, expressionless,hairs between eyes, white hair,pleated skirt,((disreveled hair)),
Negative prompt:
nsfw, owres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry,missing fingers,bad hands,missing arms, long neck, Humpbacked
其它参数:
Steps: 50
Sampler: Euler a
CFG scale: 12
Size: 960 x 640
博主实践与心得:
这份TAG生成的图像风格非常一致,大概和hiten_1有关,这也意味着采用的hiten样本中也大多是这些TAG,效果上堪比GAN,而多样性却远胜于GAN,Stable Diffusion真的未来可期。不过这些TAG中不少直接限定了具体的风格,也导致人物、背景看多了有点腻,故博主进行了下文3、中的尝试。
3、星空下的少女
博主根据2、自行删减修改得到的TAG,出图风格泛了许多,人物的发型也不再是呆毛如钳了hhh
TAG:
Positive prompt:
((masterpiece)), (((best quality))), ((ultra-detailed)), ((illustration)), beautiful detailed eyes, (1gir), (solo), beautiful detailed sky, beautiful detailed eyes, hairs between eyes, ((full body shot)), cinematic lighting, dynamic angle,
Negative prompt:
nsfw, owres, bad anatomy, bad hands, text, error, (missing fingers), extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, long neck, Humpbacked, bad crotch, bad crotch seam, fused crotch, fused seam, poorly drawn crotch, poorly drawn crotch seam, bad thigh gap, missing thigh gap, fused thigh gap, bad anatomy, short arm, (((missing arms))), missing thighs, missing calf, mutation, duplicate, more than 1 left hand, more than 1 right hand, deformed, (blurry), missing legs, extra arms, extra thighs, more than 2 thighs, extra calf, fused calf, extra legs, bad knee, extra knee, more than 2 legs,
其它参数:
Steps: 50
Sampler: Euler a
CFG scale: 9
Size: 960 x 640
博主实践与心得:
这份TAG稳定性不是很好,而且明明添加了full body shot却只出了半身像……可以尝试把负面TAG中的强调符去掉后再做尝试。
4、【NovelAI大魔导书】教科书级的壁纸展示
来自夜歩空的个人空间_哔哩哔哩_Bilibili的分享,很多小up主都从这里得以启发,视频链接:
【NovelAI大魔导书】教科书级的壁纸展示_哔哩哔哩_bilibili
另外附上该up主在动态中的TAG尝试,因信息不全本文暂不尝试动态中的TAG:动态-哔哩哔哩
TAG:
Positive prompt:
((masterpiece)), (best quality), (solo), highres, extremely detailed CG wallpaper, extremely detailed figure, Amazing, one girl (inside the church and grasses) with very long black hair and hold a (luminous) flower in hands, night with bright colorful lights whith richly layered clouds and clouded moon in the detailed sky, (many) glowing black (feathers), extremely detailed eyes, finely detail, detailed face, a lot of glowing particles, (crying) and smile, (black wing), Depth of field, perspective, (sacred feeling)
Negative prompt:
nsfw, (Multiple people), lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worstquality, low quality, normal quality, jpegartifacts, signature, watermark, username, blurry, bad feet, cropped, poorly drawn hands, poorly drawn face, mutation, deformed, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, extra fingers, fewer digits, extra limbs, extra arms, extra legs, malformed limbs, fused fingers, too many fingers, long neck, cross-eyed, mutated hands, polar lowres, bad body, bad proportions, gross proportions, text, error, missing fingers, missing arms, missing legs, extra digit
其它参数:
Steps: 30~50
Sampler: Euler
CFG scale: 7~11
Size: 1280 x 640<=尺寸<=2048 x 1024 2:1<=比例<=16:9
Denoising strength: 0.7(此参数不知如何输入)
博主实践与心得:
这份TAG大多绘制了流泪的女孩,搭配星空与植物,情境非常好。把腿部隐去、侧身像也避开了如今NovelAI模型出图的两大雷区。(聪明的你可能已经想把手臂也隐去了hhh)
推荐!
5、粉毛女路人(换装人偶)的魔咒教室
来自胧雨夜的个人空间_哔哩哔哩_Bilibili的分享,作者在NovelAI官网而非本地部署运行出图,除了给出TAG外并未提及具体参数,故参数部分博主自行测试补全。另外该负面TAG可能有矛盾冲突、或无法顺利解析语义,在1060 6G上只会生成无意义的彩色图片,故替换成3、中的负面TAG。
视频链接:
【AI绘图】NovelAI 魔法咒语tag完全解析 从入门到精通的效果优化全攻略_哔哩哔哩_bilibili
在语雀中还有大量该up主使用的TAG和使用心得:
TAG:
Positive prompt:
petite, 1girl, solo, pink hair, very long hair, school uniform, happy,outdoors, flower field, excited,((beautiful detailed water, beautiful detailed sky, fluttered detailed splashs,by leonid afremov,by wadim kashin,oni,ray tracing, girl focus,Dynamic perspective, dynamic posture,rainbow eyes, cosmic eyes,sparkling,)),
Negative prompt:
【此处不使用,改为使用3、中的负面TAG】
(lowres), (((((((((((blurry))))))))))), (((((((((bad hands))))))))), (((((missing fingers)))), (((((((((((((extra digit))))))))))))), fewer digits, small hands, error, multiple limbs, bad feet, cropped, worst quality, low quality, normal quality, jpeg artifacts, bad anatomy, long nails, ((((interlocked fingers)))), milf, nsfw, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, nsfw, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, bad hands, bad anatomy, ugly, duplicate, morbid, mutilated, tranny, trans, trannsexual, mutation, deformed, long neck, bad anatomy, bad proportions, extra arms, extra legs, disfigured, more than 2 nipples, malformed, mutated, hermaphrodite, out of frame, extra limbs, missing arms, missing legs, poorly drawn hands, poorty drawn face, mutation, poorly drawn, long body, multiple breasts, cloned face, gross proportions, mutated hands, bad hands, bad feet, long neck, missing limb, malformed limbs, malformed hands, fused fingers, too many fingers, extra fingers, missing fingers, extra digit, fewer digits, mutated hands and fingers, lowres, text, error, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, text font ui, futa, yaoi,
其它参数:
Steps: 50
Sampler: 7
CFG scale: DDIM
Size: 960 x 640
博主实践与心得:
中规中矩,不过出图不是很稳定,偶尔会生成无意义的图。作者在语雀中对加TAG的教学真的很详细,较为专业,建议大家都拜读一下。
6、原神-刻晴
来自up主小李xiaolxl的TAG分享,原神玩家狂喜。但没有负面TAG和具体参数,故负面TAG、参数部分博主自行测试补全。此处负面TAG使用3、中的负面TAG。
链接如下:
TAG:
Positive prompt:
(best quality), ((masterpiece)), (highres), (an extremely delicate and beautiful), original, extremely detailed wallpaper, hair_cones, keqing_(genshin_impact), 1girl, twintails, hair_ornament, sword, dress, gloves, petals, detached_sleeves, breasts, solo, purple_hair, medium_breasts, weapon, long_hair, holding_sword, half-closed_eyes, bare_shoulders, letterboxed, holding, frills, braid, holding_weapon, very_long_hair, closed_mouth, bangs, upper_body, long_sleeves, looking_at_viewer, frilled_sleeves, choker
Negative prompt:
【此处使用3、中的负面TAG】
nsfw, (Multiple people), lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worstquality, low quality, normal quality, jpegartifacts, signature, watermark, username, blurry, bad feet, cropped, poorly drawn hands, poorly drawn face, mutation, deformed, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, extra fingers, fewer digits, extra limbs, extra arms, extra legs, malformed limbs, fused fingers, too many fingers, long neck, cross-eyed, mutated hands, polar lowres, bad body, bad proportions, gross proportions, text, error, missing fingers, missing arms, missing legs, extra digit
其它参数:
Steps: 50
Sampler: DDIM
CFG scale: 7
Size: 960 x 640
博主实践与心得:
明显作者是研究过TAG对原神角色的刻画的,生成的也较为稳定,效果良好!同链接中还有胡桃、妮露、雷电将军、莹、甘雨、心海、宵宫等角色的TAG,快动起手来吧!
7、战场女武神
来自逝者亡魂-的贴吧的分享,非常详细的给出了加TAG炼丹的过程,包括TAG添加与参数设置技巧,非常有参考价值:
【图片】一次从零开始使用tag绘制具有氛围感和故事性画面的过程【novelai吧】_百度贴吧
TAG:
Positive prompt:
((masterpiece)), (((best quality))), ((ultra-detailed)), ((illustration)), (beautiful detailed girl), (solo), white hair, blue eyes, elf, messy_hair, ((bloodstain)), ((dirty face)), pantyhose, earrings, cape, ((knight)), battle damage, torn, (fighting_stance), falling leaves, birds, (flag), ruins, nature, wind, sunset, sunlight, beautiful detailed glow
Negative prompt:
nsfw, watermark, longbody, lowres, bad anatomy, bad hands, missing fingers, pubic hair,extra digit, fewer digits, cropped, worst quality, low quality, ((missing arms)), ((missing legs)), (((extra arms))), (((extra legs))), out of frame, gross proportions, bad feet
其它参数:
Steps: 30
Sampler: DDIM【原贴中没有,博主自己调的】
CFG scale: 7
Size: 1024 x 1024
博主实践与心得:
不愧高手在贴吧,最近绘画吧、novelai吧、AI绘画吧都有不少有意义有深度的讨论,也有不少大佬毫不吝惜的进行技术分享,盛况空前。此TAG带来的战场女武神画风一改普通二次元的画风,增加了不少深度,生成效果稳定(手臂缺失甚至可以圆了),挺有魔幻异世界的风格。
8、100+TAG画出的粉毛
来自纷纷文化的个人空间_哔哩哔哩_Bilibili的分享,魔咒之源,当初贴吧里流传的“真成咏唱魔术了,咏唱越长魔术越强”指的就是这个:
原理其实就是在扩散模型运行过程中增加生成负面图像的阻力,只要Scale较大即可严格执行prompt,代价是NLP部分比少量词条需要的GPU显存更多(部分框架、部分显卡可能撑不住一次性输入这么多的prompt,会导致生成无意义的图)。另外大量的负面TAG也要考虑不能内部冲突,不然也会生成无意义的图。
视频链接:
一位ai魔法师已经吧novelai史诗级禁咒玩明白了!完美级复刻_哔哩哔哩_bilibili
TAG:
Positive prompt:
masterpiece, best quality, best quality, Amazing, beautiful detailed eyes, 1girl, finely detail, Depth of field, extremely detailed CG unity 8k wallpaper, full body, (vtuber minato aqua), pink hair, blue streaked hair
Negative prompt:
multiple breasts, (mutated hands and fingers:1.5 ), (long body :1.3), (mutation, poorly drawn :1.2) , black-white, bad anatomy, liquid body, liquid tongue, disfigured, malformed, mutated, anatomical nonsense, text font ui, error, malformed hands, long neck, blurred, lowers, lowres, bad anatomy, bad proportions, bad shadow, uncoordinated body, unnatural body, fused breasts, bad breasts, huge breasts, poorly drawn breasts, extra breasts, liquid breasts, heavy breasts, missing breasts, huge haunch, huge thighs, huge calf, bad hands, fused hand, missing hand, disappearing arms, disappearing thigh, disappearing calf, disappearing legs, fused ears, bad ears, poorly drawn ears, extra ears, liquid ears, heavy ears, missing ears, fused animal ears, bad animal ears, poorly drawn animal ears, extra animal ears, liquid animal ears, heavy animal ears, missing animal ears, text, ui, error, missing fingers, missing limb, fused fingers, one hand with more than 5 fingers, one hand with less than 5 fingers, one hand owith more than 5 digit, one hand with less than 5 digit, extra digit, fewer digits, fused digit, missing digit, bad digit, liquid digit, colorful tongue, black tongue, cropped, watermark, username, blurry, JPEG artifacts, signature, 3D, 3D game, 3D game scene, 3D character, malformed feet, extra feet, bad feet, poorly drawn feet, fused feet, missing feet, extra shoes, bad shoes, fused shoes, more than two shoes, poorly drawn shoes, bad gloves, poorly drawn gloves, fused gloves, bad cum, poorly drawn cum, fused cum, bad hairs, poorly drawn hairs, fused hairs, big muscles, ugly, bad face, fused face, poorly drawn face, cloned face, big face, long face, bad eyes, fused eyes poorly drawn eyes, extra eyes, malformed limbs, more than 2 nipples
其它参数:
Steps: 40
Sampler: Euler
CFG scale: 12
Size: 1024 x 768
博主实践与心得:
成图率与爱丽丝非常接近,效果拔群,甚至手脚缺失的问题比爱丽丝要少!这个负面TAG的量居然没有导致生成无意义的图片,果然是值得通用的魔咒之源。
9、月色下的猫耳女孩
来自twinkleGe的个人空间_哔哩哔哩_Bilibili的分享,这位up主参考3、进行了自己的创作:
TAG:
Positive prompt:
((masterpiece)), (best quality), (highres), solo, (city), a girl with crystal yellow eyes and animal ear surrounded by (many) (feathers) (in tears), night with bright colorful lights with many clouds and clouded moon in the detailed sky, (a lot of glowing particles), medium_breast
Negative prompt:
nsfw, (Multiple people), lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worstquality, low quality, normal quality, jpegartifacts, signature, watermark, username, blurry, bad feet, cropped, poorly drawn hands, poorly drawn face, mutation, deformed, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, extra fingers, fewer digits, extra limbs, extra arms, extra legs, malformed limbs, fused fingers, too many fingers, long neck, cross-eyed, mutated hands, polar lowres, bad body, bad proportions, gross proportions, text, error, missing fingers, missing arms, missing legs, extra digit
其它参数:
Steps: 30~50
Sampler: Euler
CFG scale: 7~11
Size: 1280 x 640<=尺寸<=2048 x 1024 2:1<=比例<=16:9
博主实践与心得:
看了这么多TAG,想必大家也发现((masterpiece)), (best quality), (highres), solo是使用要诀了,网上大家还认为:
1、使用要诀:
best quality, masterpiece, highres, original, extremely detailed wallpaper
强调方式
best quality, masterpiece, highres, original, extremely detailed wallpaper, 1girl最重要的一句
an extremely delicate and beautiful:2、额外风格:
(1)草图 线稿相关:
sketch可以让图片看起来像随手画的草稿
lineart可以让线条变得很粗
posing sketch, monochrome 黑白草图
rough sketch 上了颜色的草图
monochrome+lineart情况下一般只会让眼睛上色,强调发色后头发也可以上色
monochrome, gray scale, pencil sketch lines做出的铅笔速写的感觉
利用sketch,pastel color,lineart的tag模拟一张图的绘画过程
(2)艺术风格:
chibi可以画出低头身比的效果(二头身, 三头身)
watercolor pencil可以生成彩铅画
faux traditional media可以做出签绘的风格
anime screeshot,可以让画面变成动画风格
retro artstyle赛璐璐风
photorealistic, painting, realistic, sketch, oil painting 厚涂
pastel color和sketch搭配会有速涂的质感(3)杂志 设定集风格:
official art 变得更加官方一点
three views from front, back and side和costume setup materials可以用来生成设定图
multiple views 会出现类似设定图
character sheet会出现设定图
magazine cover会把背景换成杂志封面, 配合office art更像真实杂志(虽然字没法看)
magazine scan类似杂志内页的风格
posing会强调有一个动作, 不至于出现混乱的动作(露出有六个手指头的手)
caustics 画面向主题聚焦, 类似海报
10、机甲女孩
来自的分享,NovelAI画星空、二次元女孩画的真的好!链接如下:
TAG:
Positive prompt:
(((extremely detailed CG unity 8k wallpaper)), ((masterpiece)), ((best quality)), ((one gir)), (beautiful detailed eyes), (((extreme detailed illustration))), cinematic lighting, art of light novel, highres, original, extremely detailed wallpaper, shadow , (sunshine, extreme detailed blue sky :1.1), solo, flat_chest, detailed face, night with bright colorful lights whith richly layered clouds and clouded moon in the detailed sky, high ponytail, mecha clothes, robot girl, cool movement, sliver bodysuit , (filigree), delicate and (intricate) hair, blue streaked hair, ful body, side body, Depth of field, mechanical wing in her back, flying, roof,
Negative prompt:
multiple breasts, (mutated hands and fingers :1.5), (long body :1.3), (mutation, poorly drawn :1.2), black-white, bad anatomy, liquid body, liquid tongue, disfigured, malformed, mutated, anatomical nonsense, text font ui, error, malformed hands, long neck, blurred, lowers, lowres, bad anatomy, bad proportions, bad shadow, uncoordinated body, unnatural body, fused breasts, bad breasts, huge breasts, poorly drawn breasts, extra breasts, liquid breasts, heavy breasts, missing breasts, huge haunch, huge thighs, huge calf, bad hands, fused hand, missing hand, disappearing arms, disappearing thigh, disappearing calf, disappearing legs, fused ears, bad ears, poorly drawn ears, extra ears, liquid ears, heavy ears, missing ears, fused animal ears, bad animal ears, poorly drawn animal ears, extra animal ears, liquid animal ears, heavy animal ears, missing animal ears, text, ui, error, missing fingers, missing limb, fused fingers, one hand with more than 5 fingers, one hand with less than 5 fingers, one hand with more than 5digit, one hand with less than 5 digit, extra digit, fewer digits, fused digit, missing digit, bad digit, liquid digit, colorful tongue, black tongue, cropped, watermark, username, blurry, JPEG artifacts, signature, 3D, 3D game, 3D game scene, 3D character, malformed feet, extra feet, bad feet, poorly drawn feet, fused feet, missing feet extra shoes, bad shoes, fused shoes, more than two shoes, poorly drawn shoes, bad gloves, poorly drawn gloves, fused gloves, bad cum, poorly drawn cum, fused cum, bad hairs, poorly drawn hairs, fused hairs, big muscles, ugly, bad face, fused face, poorly drawn face, cloned face, big face, long face, bad eyes, fused eyes poorly drawn eyes, extra eyes, malformed limbs, more than 2 nipples, missing nipples, different nipples, fused nipples, bad nipples, poorly drawn nipples, black nipples, colorful nipples, gross proportions . short arm, ((missing arms))), missing thighs, missing calf, missing legs, mutation, duplicate, morbid, mutilated, poorly drawn hands, more than 1 left hand, more than 1 right hand, deformed, (blurry), disfigured, missing legs, extra arms, extra thighs, more than 2 thighs, extra calf, fused calf, extra legs, bad knee, extra knee, more than 2 legs, bad tails, bad mouth, fused mouth, poorly drawn mouth, bad tongue, tongue within mouth, too long tongue, black tongue, big mouth, cracked mouth, bad mouth, dirty face, dirty teeth, dirty pantie, fused pantie, poorly drawn pantie, fused cloth, poorly drawn cloth, bad pantie, yelow teeth, thick lips, bad cameltoe, colorful cameltoe, bad asshole, poorly drawn asshole, fused asshole, missing asshole, bad anus, bad pussy, bad crotch, bad crotch seam, fused anus, fused pussy, fused anus, fused crotch, poorly drawn crotch, fused seam, poorly drawn anus, poorly drawn pussy, poorly drawn crotch, poorly drawn crotch seam, bad thigh gap, missing thigh gap, fused thigh gap, liquid thigh gap, poorly drawn thigh gap, poorly drawn anus, bad collarbone, fused collarbone, missing colarbone, liquid colarbone, strong girl, obesity, worst quality, low quality, normal quality, liquid tentacles, bad tentacles, poorly drawn tentacles, split tentacles, fused tentacles, missing clit, bad clit, fused clit, colorful clit, black clit, liquid clit, QR code, bar code, censored, safety panties, safety knickers, beard, furry, pony, pubic hair, mosaic, excrement, faeces, shit, futa, testis
其它参数:
Steps: 50
Sampler: Euler a
CFG scale: 10
Size: 1152 x 768
博主实践与心得:
跑了几张图,感觉背景的噪点比较多,可能正面TAG还需要微调。另外可能负面TAG过多了,偶尔会生成无意义的图或非半身像的图片(如脸部特写)。
欢迎交流讨论!
由浅入深理解latent diffusion/stable diffusion:一步一步搭建自己的stable diffusion models
前言: 关于如何使用stable diffusion的文章已经够多了,但是由浅入深探索stable diffusion models背后原理,如何在自己的科研中运用stable diffusion预训练模型的博客少之又少。本系列计划写5篇文章,和读者一起遨游diffusion models的世界!本文主要介绍带大家一步步搭建自己的stable diffusion models。
目录
以上是关于要点初见:Stable Diffusion NovelAI模型优质文字Tag汇总与实践魔咒汇总的主要内容,如果未能解决你的问题,请参考以下文章
ChatGPT 之后,再玩玩 Stable-Diffusion
Mac | iOS | Windows:安装Stable diffusion教程
无需本地部署 在线使用Stable Diffusion Webui 使用共享模型