Sign In

CuteYukiMix(特化可爱风格adorable style)

12k
143k
891
Updated: Dec 17, 2023
base modelanimecomic art style
Verified:
SafeTensor
Type
Checkpoint Trained
Stats
18,873
Reviews
Published
Oct 23, 2023
Base Model
SD 1.5
Usage Tips
Clip Skip: 2
Hash
AutoV2
53D7264D1C

根据建议,该版基于specialchapter版本改造,在颜色和肢体方面均有不错的改善,语意方面表现良好。

Based on the advice, this version is a transformation based on the special chapter version, showing significant improvements in both color and body language, while also performing well in conveying meaning.

关于新版本绘图参数,建议直接复制我展示例图的参数。这次版本更新依旧是关于混合的naiv3画风,以及包含了Cuteyukimix前代版本的体型等制作的风格模型,给一位好友制作的特殊版本。总体风格还是比较讨喜,但是稳定性还存在一些欠缺,并且优于前代小版本名称存在争议,这个版本我就将其起名为自己的昵称。后续我会将上个版本与这个版本的模型从各个方面持续进行优化,大家可以期待一下

Regarding the parameters for the new version of the drawing, I suggest directly copying the parameters shown in my example image.This update continues to focus on the hybrid Naiv3 art style, incorporating Cuteyukimix's predecessor's body shapes, crafted in a style tailored for a specific friend. Overall, the style remains quite appealing, yet stability still lacks in some areas, and there's controversy over the name in comparison to the previous minor versions. I've decided to name this version after my own nickname. Moving forward, I'll continuously optimize both this and the previous version's models in various aspects. Stay tuned for further improvements!

又是许久不见!不知道大家这段时间过得如何?想必各位朋友已经听说NaiV3版本堂堂登场,在这段时间也引起了不小的波澜,在经过几轮的体验以后,感觉它是非常有趣且优秀的模型。此次更新Cuteyukimix版本算是番外中的番外,是我在尝试了朋友混合的NaiV3版本画风后,迷上了该画风的笔触与色彩,想极力把它带到1.5模型,而产生的该版本。该版本模型作为初代的测试版本,还存在一些问题,例如亮度偏高,饱和度高,以及一些手部细节问题等等,接下来我也会想办法持续对它进行改进,不过它是一次非常有趣且有意义的尝试,确实将我喜欢的画风带到了1.5模型上,也让我看到了1.5模型更多的潜力。最后祝愿大家玩的开心!

Long time no see! I hope everyone has been doing well during this period. You've probably heard about the grand entrance of the NaiV3 version, which has caused quite a stir. After several rounds of exploration, I find it to be a fascinating and outstanding model. The Cuteyukimix version, introduced in this update, is like a special edition. After experimenting with a friend's hybrid NaiV3 version art style, I became captivated by its brushstrokes and colors. I was eager to bring this style to the 1.5 model, giving rise to this version. As a preliminary test version, it does have some issues like high brightness, saturation, and minor hand details. I'll continue working on improving it, but it's a fun and intentional experiment. It successfully translates the art style I love onto the 1.5 model, revealing even more potential. Lastly, I wish everyone a joyful experience with it!

许久不见!这段时间忙于生活其他事情,也是抽出空闲能回来继续研究喜欢的模型,更新作品。新版本模型是基于Neochapter3进行了改动,对clip做了比较大的调整,语义表现能力得到了加强,并且对有些泛白的低对比度进行了一定改善,并且使脸型更接近Midchapter版本的圆圆脸,更加可爱,对特殊体型的表现能力也更加接近Midchapter系列。

Long time no see! During this period, I've been busy with other aspects of life but managed to find some time to continue my research on the models I love and update my work. The new version of the model has been modified based on Neochapter3, with significant adjustments made to CLIP, enhancing its semantic expressive ability. Additionally, improvements have been made to certain washed-out low contrast areas, making the facial features closer to the round face of the Midchapter version, which is even cuter. The model's ability to represent special body types is also more akin to the Midchapter series.

midchapter3在2的基础上扩充体型范围,虽然依旧属于特化模型,但总体更偏向于对提示词有更好的灵敏度,能达到的上限更高,常态化比起之前两个版本弱化,依旧建议使用nai的vae。

The midchapter3 model expands its body size range based on the foundation of version 2. While it still falls under the category of specialized models, it is generally more inclined towards better sensitivity to prompt words and can reach higher limits. Normalization has been weakened compared to the previous two versions. It is still recommended to use NAI's VAE.

Midchapter2版本对改变体型的素材进行重新分层调整,并且也对基础配方进行改变,尽量保证稳定的形态,以满足各种用户的需求,并且对背景单调的问题也做了一些改善。

The Midchapter2 version re-layers and adjusts the materials that change the body shape, and also changes the basic formula to ensure a stable shape as much as possible to meet the needs of various users, and also makes some improvements to the problem of monotonous background.

新版本做了体型特化处理,常态化体型更可爱小巧,对特性提示词敏感度较高,不同的权重都有不错的效果,并且对颜色饱和度也加了调整,使其颜色看起来更舒服。已更新包含VAE的版本在files。

The new version strengthens the cute features to make it more in line with the cute style, and strengthens the color contrast to make the visual effect better.Updated version containing VAE in files.

NeoChapter3版本在版本2的基础上,在1个输入层和2个输出层进行了调整,尝试修复了肢体和远景手部崩坏情况,并且增加了细节(包括脸部以及人物躯干、衣服及场景细节),以及对整体构图进行了改善,并且考虑到训练TE的因素,将clip换回训练时原始的clip,以便更好地贴合tag表现,能画出版本2一些无法表现的特性,有将三个版本做xy图,对比其细节构图变化,及肢体改善,放在示例图片最后一张。已上传没有嵌入VAE的版本在files里,您可以根据自己的喜好随意挂载或嵌入合适的VAE。

The NeoChapter3 version has been adjusted based on version 2, with modifications made to one input layer and two output layers. It attempts to fix issues with limb and distant hand distortion, while also adding more details, such as facial features, character torso, clothing, and scene details. The overall composition has been improved, taking into consideration the training factors of TE. In order to better align with tag performance, the clips have been reverted back to their original training state, allowing for the portrayal of certain characteristics that version 2 was unable to depict. Three versions are then depicted in an XY graph, comparing the changes in detailed composition and limb improvement. This is shown on the last example image.The uploaded version without VAE is in the files, you can mount or embed the appropriate VAE according to your preference.

关于NeoChapter2版本模型,增强了脸部细节,相较于上个版本脸部细节模糊情况进行了改善,整体近远景细节丰富度也有提升,迁移了clip加强了对语意的理解能力,更好的贴合tag提示词,肢体上进行了加强,并且对于手部刻画也进行了一定的修复。色调明亮度稍有提升。已将含有VAE的版本替换了files里的VAE文件,内置animeave.pt。

Regarding the NeoChapter2 version of the model, there have been enhancements made to facial details. Improvements have been made to address the issue of blurry facial details compared to the previous version. There is also an overall increase in the richness of details in both close-up and distant views. The integration of CLIP has further strengthened its understanding of semantics, resulting in better alignment with tag prompts. There have been improvements in depicting body movements and some fixes have been made to hand gestures as well.Additionally, there has been a slight increase in brightness levels.The version containing the VAE has been replaced with the VAE file in the "files" section, which is named animeave.pt.

关于新章版本,做了一些风格上的改动,您在使用它生成图像时可以感受到与之前版本的差异,扩充了人物的体型,但是保留了可爱的脸型,整体色调偏淡,如果您想要生成更鲜艳的图像,建议您配合其他vae使用,并且在后续的版本我也会对颜色持续调整以达到最好的效果,顺带一提,新版本生成的脸真的很可爱,期待您使用以后的反馈。已上传.ckpt包含vae的版本,可在files里面选择下载。

Regarding the NeoChapter version, some stylistic changes have been made. You will notice a difference in the generated images compared to the previous version. The body proportions of the characters have been expanded while retaining their cute facial features. The overall color palette leans towards lighter tones. If you prefer more vibrant images, it is recommended to use other VAEs in conjunction with this version. In future updates, I will continue to fine-tune the color settings to achieve the best results. By the way, the faces generated by the new version are really adorable. I look forward to receiving your feedback after you've tried it out.The .ckpt file has been uploaded, which contains the version of VAE. You can choose to download it from the "files" section.

关于新版本,我将在介绍页面单独为它写相关介绍,希望您可以了解它的详情以及感受到它

Regarding the new version, I will write a separate introduction for it on the introduction

的魅力。

page. I hope you can learn the details and feel its charm.

相较于之前赛璐璐以及平涂风格的模型,我对该系列模型在细节以及色调上进行了加强,您

Compared to the previous models, such as Sally and the flat painting style, I have

在使用该模型的过程中应该会感受到比以往版本更多的细节呈现,以及更加舒适的色彩。并

enhanced the series in terms of details and color tones.When using this model, you

且在画风方面较以前略显生硬的地方做了圆润化处理,使其更符合cute风格。

should experience more detailed representation and more comfortable colors than in

我相信,对于喜欢这一风格系列模型的用户来说,您们都对可爱的事物充满热爱。我希望这

previous versions. I have made the style smoother, giving it a more fitting cute

一模型能帮助您们创作出属于自己的最佳作品。同时,我也会不断对该系列模型进行优化。

aesthetic.I believe that those who appreciate this style of models have a deep love for

如果您有任何问题,请在评论区与我交流。

adorable things. I hope this model can help you create your best works that truly belong to you. Furthermore, I will continue to optimize this series of models. If you have any questions, please feel free to communicate with me in the comments section.

这个模型是私炉的产物,整体风格偏幼化。它可以比较轻松的画出可爱风格的人物,但并非只能画萝莉风格人物,您可以尝试各种有意思的prompt,并且它画风景(背景)效果也很好。
关于v2版本,增强稳定性以及细腻程度,推荐参数如下:
Steps: 20-30, Sampler: DPM++ SDE Karras, CFG scale: 7, Size: 512x768, Denoising strength: 0.6, Clip skip: 2, Hires upscale: 2, Hires steps: 10-25, Hires upscaler: R-ESRGAN 4x+ Anime6B.
Negative prompt:NG_DeepNegative_V1_75T, EasyNegative, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, (worst quality, low quality:1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit,fewer digits, (extra arms:1.2), (bad-artist:0.6), bad-image-v2-39000

The reason for this model's name is that it can easily create cute-style images, and its overall color tone is also visually appealing, making it suitable for landscape-style pictures as well. We hope you'll enjoy it!
Regarding the V2 version, to enhance stability and delicacy, the recommended parameters are as follows:

Steps: 30, Sampler: DPM++ SDE Karras, CFG scale: 7, Size: 512x768, Denoising strength: 0.57, Clip skip: 2, Hires upscale: 2, Hires steps: 25, Hires upscaler: R-ESRGAN 4x+ Anime6B.
Negative prompt:NG_DeepNegative_V1_75T, EasyNegative, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, (worst quality, low quality:1.4), (bad anatomy), (inaccurate limb:1.2), bad composition, inaccurate eyes, extra digit,fewer digits, (extra arms:1.2), (bad-artist:0.6), bad-image-v2-39000