SDXL版本的基础模型和LoRA还有存在感么？（模型介绍系列DreamShaper XL1.0-综合型）

今天跟大家介绍的这款，作者Lykon所作：DreamShaper XL1.0模型。该模型是最早一批发布在C站，基于SDXL版本训练的基础模型。

作者之前就做了很多代的基于SD1.5的模型，最新版本是dreamshaper_8。

虽说SDXL版在AI绘画方面已经具有了很强的能力的官方模型base1.0和refiner1.0模型，但经过lykon训练的这款DreamShaper在综合表现能力方面还是有一定的提升。个人感觉，这个模型在人物、场景、动物等方面对画面的解读和绘画能力相比官方的Base模型还是有不小提升，因此我会把他归类为综合型基础模型。以下是本人用部分画面来比对了两个模型的出图效果，有些主题的画面感觉差别不是太大，这也有是因为SDXL版本官方模型Base1.0本就很强大的缘故。大家也可以对比和自行下载体会。

1、画男人

photo of the warrior Aragorn from Lord of the Rings, film grain, 8k hd

Negative prompt: (deformed iris, deformed pupils), text, worst quality, low quality, ugly

Steps: 40, Seed: 17748028598464, Model: DreamShaperXL1.0Alpha_half, width: 768, height: 1024, Sampler: DPM++ 2S a Karras, CFG scale: 8

2、画女人

(masterpiece), (extremely intricate:1.3), (realistic), photo of a girl, the most beautiful in the world, (medieval armor), upper body, outdoors, far away castle, professional photograph of a stunning woman detailed, sharp focus, dramatic, award winning, cinematic lighting, volumetrics dtx, (film grain, blurry background, blurry foreground, bokeh, depth of field, sunset, motion blur:1.3), 4k, 8k, hd, hdr

Negative prompt: (deformed iris, deformed pupils), text, worst quality, low quality, ugly

Steps: 40, Seed: 17748028598465, Model: DreamShaperXL1.0Alpha2_fixedVae_half_00001_, width: 768, height: 1024, Sampler: DPM++ 2S a Karras, CFG scale: 8

3、画物体

ultra detailed 8k cg,japanese ramen,chopsticks,egg,steam,

Negative prompt: embedding:BadDream,embedding:UnrealisticDream,

Steps: 40, Sampler: DPM++ 2M SDE Karras, CFG scale: 8, Seed: 17748028598465, Size: 1280x968,

4、画动物

RAW photo,Fuji X-T20,high quality,a baby lion,cute,

Negative prompt: 3d render,anime,Signature,deformed,3d render,smooth,plastic,blurry,grainy,anime,embedding:BadDream embedding:UnrealisticDream,

Steps: 30, Sampler: DPM++ 2S a Karras, CFG scale: 8, Seed: 433200653537535,

5、画场景1

(breathtaking:1.25),(watercolor, aquarelle, wet watercolor:1.25),by alena aenami,

(intricate details),(hyperdetailed),8k hdr,high detailed,lot of details,high quality,

Negative prompt: dof,blur,bokeh,monochrome,

Steps: 25, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 3038583188, Size: 1280x968, Model hash: 0f1b80cfe8, Model: dreamshaperXL10_alpha2Xl10, Style Selector Enabled: True, Style Selector Randomize: False, Style Selector Style: base, Version: v1.5.2

6、画场景2

macabre style of a tiny red flower growing on the street on an abandoned city. Nadir image. dark,gothic,grim,haunting,highly detailed,

Negative prompt: bright,cheerful,light-hearted,cartoonish,cute,

Steps: 30, Sampler: DPM++ 2M SDE Karras, CFG scale: 8, Seed: 3943704579, Size: 1280x968, Model hash: 0f1b80cfe8, Model: dreamshaperXL10_alpha2Xl10, Style Selector Enabled: True, Style Selector Randomize: False, Style Selector Style: base, Version: v1.5.2

7、产品图

product shot of a computer tower with a glass cover,the inside is (on fire:1.75),breathtaking,highly detailed,award-winning,

Negative prompt: anime,cartoon,graphic,text,painting,crayon,graphite,abstract glitch,blurry,

Steps: 40, Sampler: DPM++ 2M SDE Karras, CFG scale: 5, Seed: 1881540143, Size: 1280x968, Model hash: 0f1b80cfe8, Model: dreamshaperXL10_alpha2Xl10, Clip skip: 2, Style Selector Enabled: True, Style Selector Randomize: False, Style Selector Style: base, Version: v1.5.2

8、配合LoRA

full body complex render ultra detailed of a beautiful porcelain cracked (adorable:1.3) android dachshund,cyborg,robotic parts,150mm,beautiful studio soft light,rim light,vibrant details,luxurious cyberpunk,hyperrealistic,anatomical,cable electric wires,microchip,elegant,beautiful background,octane render,iridescent silver and brown,masterwork by head of prompt engineering,lora:wowifierXL:0.2,art by mooncryptowow,

Negative prompt: boring,nude,text,signature,logo,watermark,

Steps: 32, Sampler: DPM++ 2M SDE Karras, CFG scale: 7, Seed: 2760689376, Size: 1024x768, Model hash: 0f1b80cfe8, Model: dreamshaperXL10_alpha2Xl10, Denoising strength: 0.25, Style Selector Enabled: True, Style Selector Randomize: False, Style Selector Style: base, Hires upscale: 2, Hires steps: 20, Hires upscaler: R-ESRGAN 4x+, Lora hashes: “wowifierXL: 45866502260d”, Version: v1.5.2

9、抽象概念

(dark magic:1.2),(grim:1.4),(redscale, woodcut:1.2),(etching:1.1),(engraving:0.4),(Magical Photo:1.3) of a eclectic (dark shot:1.2),80mm,(dark magic:1.2),(grim:1.4),glossy glass ball with (glowing small pine, dramatic lighting intricate, vibrant colors, cinematic) inside,surrounded by stones,green plants and mossy ground cover in the woods,a macro photograph,(hdr:1.2),complex background,hyperdetailed,art [[by jordan grimmer]],(intricate details),(hyperdetailed),8k hdr,soft cinematic light,dramatic atmosphere,atmospheric perspective,by Ali Erturk,(intricate details:0.9),(hdr, hyperdetailed:1.2),soft light,sharp,exposure blend,medium shot,bokeh,(hdr:1.3),high contrast,(cinematic:0.9),(muted colors, dim colors, soothing tones:1.15),(hyperdetailed:1.2),(noir:0.4),strobist,(intricate details),(hyperdetailed),8k hdr,soft cinematic light,dramatic atmosphere,atmospheric perspective,

Negative prompt: blur,blurry,text,error,signature,watermark,logo,bw,grayscale,dof,neon,blur,blurry,text,error,signature,watermark,logo,(deformed, distorted, disfigured:1.3),poorly drawn,bad anatomy,wrong anatomy,extra limb,missing limb,floating limbs,(mutated hands and fingers:1.4),disconnected limbs,mutation,mutated,ugly,disgusting,blurry,amputation,over saturated,

Steps: 75, Sampler: DPM++ 2M SDE Karras, CFG scale: 5.5, Seed: 1993573135, Size: 1024x768, Model hash: 0f1b80cfe8, Model: dreamshaperXL10_alpha2Xl10, Denoising strength: 0.7, Style Selector Enabled: True, Style Selector Randomize: False, Style Selector Style: base, Hires upscale: 2, Hires steps: 8, Hires upscaler: R-ESRGAN 4x+, Version: v1.5.2

=========================

模型下载百度云盘：

https://pan.baidu.com/s/1_j7g_AkR7ulU1eaV13c81A?pwd=1111

加入我们：

欢迎关注我们公众号，一起研习：

欢迎扫码入群，与我们共同探讨：

关于AI绘画技术储备

学好 AI绘画 不论是就业还是做副业赚钱都不错，但要学会 AI绘画 还是要有一个学习规划。最后大家分享一份全套的 AI绘画 学习资料，给那些想学习 AI绘画 的小伙伴们一点帮助！

👉[[CSDN大礼包：《StableDiffusion安装包&AI绘画入门学习资料》免费分享]]（安全链接，放心点击）

1.stable diffusion安装包

随着技术的迭代，目前 Stable Diffusion 已经能够生成非常艺术化的图片了，完全有赶超人类的架势，已经有不少工作被这类服务替代，比如制作一个 logo 图片，画一张虚拟老婆照片，画质堪比相机。

最新 Stable Diffusion 除了有win多个版本，就算说底端的显卡也能玩了哦！此外还带来了Mac版本，仅支持macOS 12.3或更高版本。

2.stable diffusion视频合集

我们在学习的时候，往往书籍源码难以理解，阅读困难，这时候视频教程教程是就很适合了，生动形象加上案例实战，一步步带你入坑stable diffusion，科学有趣才能更方便的学习下去。

3.stable diffusion模型下载

stable diffusion往往一开始使用时图片等无法达到理想的生成效果，这时则需要通过使用大量训练数据，调整模型的超参数（如学习率、训练轮数、模型大小等），可以使得模型更好地适应数据集，并生成更加真实、准确、高质量的图像。

4.stable diffusion提示词

提示词是构建由文本到图像模型解释和理解的单词的过程。可以把它理解为你告诉 AI 模型要画什么而需要说的语言，整个SD学习过程中都离不开这本提示词手册。

5.SD从0到落地实战演练

如果你能在15天内完成所有的任务，那你堪称天才。然而，如果你能完成 60-70% 的内容，你就已经开始具备成为一名SD大神的正确特征了。

这份完整版的学习资料已经上传CSDN，朋友们如果需要可以微信扫描下方CSDN官方认证二维码免费领取【保证100%免费】