今天推荐一个非常优秀的AI动漫大模型，这个模型的画面质量堪比手绘，从效果看，几乎可以接商单了。
模型介绍 ：
这个基于Laxhar/noobai-XL_v1.0的图像生成模型，利用了完整的Danbooru和e621数据集，包括原生标签和自然语言描述。
作为一个v-预测模型（与eps-预测模型不同），它需要特定的参数配置——在接下来的部分中有详细说明。
特别感谢我的队友euge的编码工作，同时我们对许多提供技术支持的社区成员表示感激。
⚠️ 重要通知 ⚠️ 此模型的工作方式与EPS模型不同！请仔细阅读指南！
模型详情 开发者：Laxhar实验室
模型类型：基于扩散的文本到图像生成模型
从以下模型微调而来：Laxhar/noobai-XL_v1.0
赞助商：
Lanyun Cloud
Civitai & Seaart
协作测试：
LiblibAI
Nieta
如何使用模型。NoobAI XL的使用指南：英文：
https://civitai.com/articles/8962
中文：
https://fcnk27d6mpa5.feishu.cn/wiki/S8Z4wy7fSiePNRksiBXcyrUenOh
方法I：reForge （如果您尚未安装reForge）请按照仓库中的说明安装reForge；
启动WebUI并像平常一样使用模型！
方法II：ComfyUI SAMLPLE与NODES
comfy_ui_workflow_sample
方法III：WebUI 请注意dev分支不稳定，可能包含错误。
（如果您尚未安装WebUI）请按照仓库中的说明安装WebUI。为了简化
切换到dev分支：
git switch dev 3. 拉取最新更新：
git pull 4. 启动WebUI并像平常一样使用模型！
方法IV：Diffusers import torch from diffusers import StableDiffusionXLPipeline from diffusers import EulerDiscreteScheduler
ckpt_path = "/path/to/model.safetensors" pipe = StableDiffusionXLPipeline.from_single_file( ckpt_path, use_safetensors=True, torch_dtype=torch.float16, ) scheduler_args = {"prediction_type": "v_prediction", "rescale_betas_zero_snr": True} pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, **scheduler_args) pipe.enable_xformers_memory_efficient_attention() pipe = pipe.to("cuda")
prompt = """masterpiece, best quality,artist:john_kafka,artist:nixeu,artist:quasarcake, chromatic aberration, film grain, horror themetheme, limited palette, x-shaped pupils, high contrast, color contrast, cold colors, arlecchino genshinimpactgenshinimpact, black theme, gritty, graphite mediummedium""" negative_prompt = "nsfw, worst quality, old, early, low quality, lowres, signature, username, logo, bad hands, mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro"
image = pipe( prompt=prompt, negative_prompt=negative_prompt, width=832, height=1216, num_inference_steps=28, guidance_scale=5, generator=torch.Generator().manual_seed(42), ).images[0]
image.save("output.png") 注意：请确保您的机器上已安装Git，并且环境配置正确。
推荐设置 参数 CFG：4 ~ 5
步骤：28 ~ 35
采样方法：欧拉（⚠️ 其他采样器可能无法正常工作）
分辨率：总面积约为1024x1024。最佳选择：768x1344, 832x1216, 896x1152, 1024x1024, 1152x896, 1216x832, 1344x768
提示 提示前缀：
masterpiece, best quality, newest, absurdres, highres, safe, 负面提示：
nsfw, worst quality, old, early, low quality, lowres, signature, username, logo, bad hands, mutated hands, mammal, anthro, furry, ambiguous form, feral, semi-anthro 使用指南 标题 <1girl/1boy/1other/...>, <character>, <series>, <artists>, <special tags>, <general tags>, <other tags> 质量标签 对于质量标签，我们通过以下流程评估图像的受欢迎程度：
基于不同来源和评级的数据归一化。
根据日期的时效性应用衰减系数。
基于此处理在整个数据集中对图像进行排名。
我们的最终目标是确保质量标签有效地追踪近年来用户偏好。
百分位范围质量标签> 95thmasterpiece> 85th, <= 95thbest quality> 60th, <= 85thgood quality> 30th, <= 60thnormal quality<= 30thworst quality
审美标签 标签描述very awaTop 5%的图像在审美评分方面由waifu-scorer最佳审美All底部5%的图像在审美评分方面由waifu-scorer和aesthetic-shadow-v2......
日期标签 有两种类型的日期标签：年份标签和时期标签。对于年份标签，请使用年份xxxx格式，即年份2021。对于时期标签，请参考以下表格：
年份范围时期标签2005-2010old2011-2014early2014-2017mid2018-2020recent2021-2024newest
数据集 训练日期之前的最新Danbooru图像（大约在2024-10-23之前）
E621图像e621-2024-webp-4Mpixel数据集在Hugging Face上
沟通
QQ群：
875042008
914818692
635772191
870086562
Discord: Laxhar Dream Lab SDXL NOOB
如何在v-pred SDXL模型上训练LoRA
针对基于sd-scripts的LoRA训练者的教程。
文章链接：https://civitai.com/articles/8723
实用工具
Laxhar实验室正在为NoobXL训练一个专门的ControlNet模型，并且正在逐步发布模型。到目前为止，已经发布了正常、深度和canny。
模型链接：https://civitai.com/models/929685
模型许可 此模型的许可继承自https://huggingface.co/OnomaAIResearch/Illustrious-xl-early-release-v0 fair-ai-public-license-1.0-sd，并添加了以下条款。任何使用此模型及其变体的行为都受此许可约束。
I. 使用限制 禁止将模型用于有害、恶意或非法活动，包括但不限于骚扰、威胁和传播错误信息。
禁止生成不道德或冒犯性内容。
禁止违反用户所在管辖区的法律和法规。
II. 商业禁止 我们禁止任何形式的商业化，包括但不限于模型、衍生模型或模型生成产品的货币化或商业使用。
III. 开源社区 为了培养一个蓬勃发展的开源社区，用户必须遵守以下要求：
开源衍生模型、合并模型、LoRAs以及基于上述模型的产品。
分享工作细节，如合成公式、提示和工作流程。
遵循fair-ai-public-license以确保衍生作品保持开源。
IV. 免责声明 生成的模型可能会产生意外或有害的输出。用户必须承担使用的所有风险和潜在后果。
参与者和贡献者 参与者 L_A_X: Civitai | Liblib.art | Huggingface
li_li: Civitai | Huggingface
nebulae: Civitai | Huggingface
Chenkin: Civitai | Huggingface
Euge: Civitai | Huggingface | Github
贡献者 Narugo1992: 感谢narugo1992和deepghs团队开源各种训练集、图像处理工具和模型。
Mikubill: 感谢Mikubill提供的Naifu训练器。
Onommai: 感谢OnommAI开源一个强大的基础模型。
V-Prediction: 感谢以下个人提供的详细指导和实验。
adsfssdf
bluvoll
bvhari
catboxanon
parsee-mizuhashi
very-aesthetic
momoura
madmanfourohfour
社区：aria1th261, neggles, sdtana, chewing, irldoggo, reoe, kblueleaf, Yidhar, ageless, 白玲可, Creeper, KaerMorh, 吟游诗人, SeASnAkE, zwh20081, Wenaka~喵, 稀里哗啦, 幸运二副, 昨日の約, 445, EBIX, Sopp, Y_X, Minthybasis, Rakosz, 孤辰NULL, 汤人烂, 沅月弯刀, David
1、动漫海报
提示词：
1girl,,eula \(genshin impact\),,::, (by kana616:0.8), [(by hen-tie:1.3)|(by kagami_\(galgamesion\):1.2)], (by kinokohime:1.2), (by yatsuha_\(hachiyoh\):0.8), year 2024,,(gradient:1.40), (close-up:0.80), (turtleneck:1.20), (from side:1.20), (wide sleeves:1.20), (monochrome:1.20), (belt:1.20), (copyright name:1.20),,from side, hand on headwear, closed mouth, flat color,,medium breasts, medium hair, large breasts, black hairband, breasts, blue hair, hair ornament, gloves, hairband,,blue gloves, crescent hat ornament, skindentation, witch hat, thigh strap, star hat ornament, hat ornament, elbow gloves, pantyhose, star print,,red hat, blue hat,,day, blurry background, sky, cloudy sky, on grass, transformer, building, gradient, blurry, utility pole, outdoors, cloud, blue sky, grass,sunlight, colorful, very awa, masterpiece, best quality, newest, highres, absurdres
2、明日香
提示词：
masterpiece, best quality, amazing quality, good quality, very aesthetic, absurdres, highres, by (lifeline \(a384079959\):0.6) , (dsmile:0.6) and (lack:0.6),
break
1girl, Asuka Langley Soryu, neon genesis evangelion, rebuild of evangelion, sitting on the edge of a city rooftop at twilight, overlooking a sprawling cityscape. She wears a red leather jacket, ripped jeans, and combat boots, exuding a rebellious vibe. Her auburn hair flows in the breeze as she leans back with a confident smirk. The city below is bathed in purple and orange hues, with skyscrapers reflecting the fading light. The composition captures her bold personality, framed by the dramatic skyline. from behind, looking back, from below, even framing, fullbody shot, mirrored background, ruffled clothing, looking at viewer. wind effect, floating long hair,
break
High-Definition Anime Character Detailing, Polished Anime Illustration, Japanese Artistic Style with Enhanced Clarity, Anime Ultra HD, Anime Style with Realistic Textures, Crisp Anime Lines, Vivid Anime Colors, Rich Color Palette in Anime, Smooth Shading in Anime Style, Anime with Depth of Field Effect, high contrast, Professional fashion editorial feel, ethereal, highly detailed, ultra detailed, insanely detailed, fine texture, crisp detail, crisp edges, [3d:cel shading:0.6], depth of field, bokeh,
3、你头上戴着一朵花，等着有人带你去流浪
提示词：
masterpiece, best quality, good quality, very aesthetic, absurdres, newest, 1girl, solo, long hair, looking at viewer, black hair, hair ornament, red eyes, dress, closed mouth, jewelry, upper body, flower, earrings, artist name, hair flower, blurry, from side, parted bangs, eyelashes, sparkle, petals, blurry background, makeup, depth of field, expressionless, rose, chinese clothes, red dress, portrait, red flower, gem, lipstick, china dress, light particles, eyeshadow, red rose, glint, red lips, falling petals, red theme, red eyeshadow, red gemstone, long eyelashes
4、伸出你的手
提示词：
masterpiece, best quality,highres,newest,safe,1girl,hatsukoi_jikan_-_anri, long_hair,head tilt,sliver hair, hair_ribbon, side_ponytail,medium breasts,purple eyes,wheat_field,outdoors,pink sundress,collarbone,(sleeveless:1.2),hand behind back,(outstreched arm,reaching_towards_viewer :1.3),dutch angle,smile,teeth,blush,solo
5、姐妹
提示词：
very awa, masterpiece, newest BREAK 2girls, prinz eugen \(azur lane\), azur lane, multiple girls, sitting, side-by-side, on couch, holding cup, arm at side, parted lips, smile, looking at viewer, mother and daughter, mature female, large breasts, medium breasts, alternate hairstyle, alternate costume, matching outfits, turtleneck sweater, white sweater, pantyhose, hair ribbon, black ribbon, steam, coffee mug, window, snow, snowing, indoors, cold lighting
6、蒙住了眼睛
提示词：
masterpiece, best quality, good quality, very aesthetic, absurdres, newest,1girl, solo, hanfu,long hair, black hair, long sleeves, closed mouth, jewelry, ponytail, choker, hand up, grey background, black hanfu, bracelet, torn clothes, White sash, black choker, chinese clothes, facing viewer, Black bandages over eyes, red lips, hand to own mouth, covered eyes, wind, sunlight
7、飞远的精灵
提示词：
rella ,1girl, full body ,glitch , (colorful:0.8),one arm behind back,butterfly wings,dynamic pose,(waving at viewer),5fingers looking back,smile, white dress,(leaning forward:0.8), bare feet,floating,(gradient background:1.2) ,(chromatic_aberration:1), light particles,night,[(digital dissolve,dissolving :1):(digital dissolve,dissolving :1.3):15], masterpiece,best quality,year 2024,newest,highres,absurdres,very Aesthetic
8、星河下的蕾姆
提示词：
artist:miv4t, (artist:mika pikazo:0.7), (artist:neko no kedama), [artist:lack],
1girl, solo, rem \(re:zero\), re:zero kara hajimeru isekai seikatsu, maid, maid, maid headdress, blue eyes, hair over one eye, blue hair, short hair, x hair ornament, hair ribbon, medium breasts, large breasts, cleavage, roswaal mansion maid uniform, roswaal mansion maid uniform, detached sleeves, smile, open mouth, full body, spread arms, (from afar:1.3), dim, dark, outdoors, star \(sky\), galaxy, grass, hill
((masterpiece, best quality, absurdres, highres, newest, very aesthetic, extremely detailed, very awa, year 2024, detailed background, perfect anatomy, surreal))
网盘下载：
https://pan.quark.cn/s/fde6cad6b6ab
