图图在自己的社群里举办了：赵图图的视觉奇迹挑战活动，这是一个长期活动，只要在群里分享优质绘图作品，就有可能获得独特社区称号+算力+私模+神秘礼品+全渠道宣传+交流答疑，628266084

电报：https://t.me/+nbU3j7rLEYZkNzVl

I found a new AI tool Shakker, a best image to image tool. You can try it via https://www.shakker.ai

V3.0

Communication Q group: 950351015, welcome to play and exchange information.

V3.0

Compared with the previous version, it has made significant progress, mainly in the aspects of less pollution, good generalization, and ease of use.

Brand: Moumi

Series: Aurora Pure Enjoyment 2.0

Material: 8D 528N

Color: black

On the basis of Difference Extraction Method 3.0, the production method has been improved to a certain extent and produced with real shots.

For ease of use, all the LORA trigger words I create in the future will be tutututu, four tu.

For this LORA, the trigger word is:

tutututu,black pantyhose,

All large models are friendly.

Instructions:

1. Make sure you use LORA correctly.

2. It is recommended that the number of CLIP termination layers be set to 2.

3. The weight depends on the needs. It can be from 0 to 1. It is recommended to start with 0.6.

4. The trigger words are: tutututu, black pantyhose, (if the crotch is open, add the prompt word Crotchless pantyhose)

5. Don’t add too many negatives, otherwise it will affect the effect of LORA.

6. The prompt word guidance coefficient (CFGScale) can be tried starting from 7.

7. I like to use DDIM as the sampling method, but the other ones are actually pretty good too. Please try more parameters~

If you encounter problems during use, you can leave me a message.

Looking forward to your work!

Looking forward to your work!

Looking forward to your work!

Hope you have a good time.

Thank you for your long-term trust and support. If you have any questions, please give us feedback and we will continue to optimize them later~

statement:

1. You shall bear full responsibility for any creative works using this model.

2. You may not use this model to intentionally create or share illegal or harmful output or content, and avoid using this model for malicious, harmful, defamatory, fraudulent or political purposes.

3. If you use this model for commercial purposes, please inform me, thank you.

common problem:

1. What is differential extraction method?

As the name suggests, simply speaking, it is to extract the differences to make LORA. Our general LORA is trained directly with pictures. In this way, the trained LORA will inevitably be contaminated by the base film and other environmental factors, so it will have an unexpected impact on the picture when the picture is produced. The LORA produced by the differential extraction method will remove these contaminations. In this way, even if the weight of the LORA produced is set relatively high, the pollution to the picture will be very small, and the model quality will be better. The differential extraction method is more complex to make and requires more steps.

2. Why do you want to do this LORA?

Although there are many similar LORAs, I found that on the one hand, most of these LORAs are not very close to reality. Many of the more popular clothing at the moment are not available, and they are not realistic enough.

In addition, the quality is uneven. Many LORAs have a greater impact on the characters themselves and the graphics, making them inconvenient to use. So I plan to make a series of clothing LORA.

3. Why can’t I take photos like yours?

Everyone's computer configuration and operating environment are different, and the LORA and plug-in versions used, plug-in configurations, etc. will also be different, so please don't pursue the same photos and create your own perfect works!

4. How to get in touch?

If you have any need to customize LORA, or have other cooperation intentions, you can leave a message here directly, or contact Q331506796 (indicate your intention)

5. What optimizations have you made to the difference extraction method? Or are you just talking casually?

Answer: I have made a lot of optimizations to the code involved:

1. Automated model exploration: Automatically generate multiple fusion models from a set of models, similar to the automated process of parameter tuning or model selection, but focused on model fusion.

2. Difference analysis based on Euclidean distance: used to compare the weight differences of two LoRA (Low-RankAdaptation) models.

In dissimilarity analysis, Euclidean distance is used to quantify the difference between the weights of two models. Specifically:

Weight representation: The weight of each LoRA model can be regarded as a point in high-dimensional space. Each element of the weight corresponds to a dimension in the space.

Calculate the difference: By calculating the Euclidean distance between the corresponding weights of the two models, we can get a numerical value that represents the relative position difference of the two weights in the high-dimensional space.

Explain the difference: A larger Euclidean distance means the two models are more different in that weight; a smaller distance means they are more similar.

This difference analysis is used to guide model fusion decisions. For example, if two models differ greatly in a certain weight, this weight may need to be specially processed to ensure that the fused model can effectively combine the characteristics of the two original models.

These optimizations are completely based on my own materials. Because my materials have some special characteristics, they are not difference extraction in the traditional sense. There are some trade-offs.

6. What is the difference between V3.0 and V2.0?

V3.0 can be said to be a brand new LORA, which can be said to be an all-round improvement.

1. Shooting

V2.0 uses SK's Aurora. In order to further improve the quality, I purchased a number of better products on the market for comparison, including Jirun K9, Jirun K1-2.0 version, Laimi Aurora Pure Enjoyment 2.0, Laimi Aurora, Laimi Dark Gold 2.0, Ling Liuguang, JM is wonderful, SK Aurora high-density version.

List of some test products

After the final test, Laimi Aurora Pure Enjoy 2.0 stood out and had the best effect.

Lai Mi Aurora Pure Enjoyment 2.0

Jirun K9

Jirun K1-2.0 version

Ling Liuguang

In order to avoid the unnatural situation of the first two versions, I made special processing on the shooting method, such as adding materials under natural light instead of lighting them all in the studio, so that the effect will be more natural. (The main thing is to avoid the same light as the line in the middle, otherwise there will be problems in the picture. This is also the key point to improve in this version.)

2. Production

In addition to using self-developed fusion algorithms during production, the data set was also deeply processed to ensure the effect. At the same time, in order to avoid the mismatch between the materials and the actual effects, I produced two sets of LORA, and after intelligent fusion and exploration, 50 models were formed.

3. Test

After excluding 25 of the samples, the following figure shows the comparison effect of the remaining 25 samples:

The test results focus on the following aspects:

1. When the weight is 1, do not modify the face, the face will not collapse, and try to maintain the original rendering effect of the large model.

2. When no descriptors of clothes are added, it will have the least impact on any picture elements other than black silk.

3. The impact of shoes and character movements is minimal, and there is no change in the overall tone of the picture.

Finally, models No. 17 and No. 24 were selected to enter the next round of testing.

4. Test and compare the rendering effects under different base films, and explore the weights, especially when a lot of LORAs are used together.

5. It can be seen that the effect of No. 24 is better, so I finally chose No. 24.

6. Compare horizontally with all other models of the same type on the entire network.

7. It costs 1,000RMB to make a model, and 700RMB to buy Hi-Silk. It’s a waste of money, and it’s disbanded.

V3.0

图图在自己的社群里举办了：赵图图的视觉奇迹挑战活动，这是一个长期活动，只要在群里分享优质绘图作品，就有可能获得独特社区称号+算力+私模+神秘礼品+全渠道宣传+交流答疑，628266084

相比之前的版本有大幅度的进步，主要还是体现在污染少，泛化好，使用方便这几个方面。

品牌：某觅

所属系列：极光纯享2.0

材质：8D 528N

颜色：black

在差异提取法3.0的基础上又对制作方法进行了一定的改良，实拍制作。

为了方便使用，以后我制作的LORA触发词全部都是tutututu,四个tu。

对于这个LORA，触发词为：

tutututu,black pantyhose,

各种大模型都很友好。

使用方法：

1、确保你正确的使用LORA.

2、CLIP终止层数建议设置为2.

3、权重看需要，从0到1都可以，建议0.6开始。

4、触发词为：tutututu, black pantyhose,（开裆的话加入提示词 Crotchless pantyhose）

5、负面不要加的太多，否则会对LORA的效果产生影响。

6、提示词引导系数(CFGScale)可以从7开始尝试。

7、采样方法我喜欢用DDIM,其他的其实也还好，参数多多尝试哈~

如果您在使用中遇到问题，可以给我留言。

期待您的作品！

期待您的作品！

期待您的作品！

祝你玩的开心。

感谢您长期以来的信任和支持。有问题请反馈，后面还会继续优化哈~

声明：

1、您应对使用此模型的任何创意作品承担全部责任。

2、您不能使用该模型故意制作或分享非法或有害的输出或内容，避免将此模型用于恶意、伤害、诽谤、诈骗或政治用途。

3、如果将本模型用于商业用途，请通知我，谢谢。

常见问题：

1、什么是差异提取法？

顾名思义，简单来说就是将差异提取出来制作LORA。我们一般的LORA都是拿图直接训练，这样训练出的LORA不可避免的会受到底膜以及其他环境因素污染，所以在出图的时候会对画面产生预料之外的影响。差异提取法制作的LORA会将这些污染去除掉，这样制作出的LORA权重即使开的比较高对画面的污染也很小，模型质量会更好。差异提取法制作起来更复杂，需要的步骤更多。

2、为什么要做这个LORA?

虽然类似的LORA有很多，但是我发现一方面这种LORA大多数不是很贴近现实，很多当下比较流行的服装并没有，而且不够写实。

另外就是质量方面残次不齐，很多LORA对人物本身以及画面的影响比较大,使用起来不够方便。所以打算做一系列的服装LORA。

3、为什么我出不了和你一样的照片？

每个人的电脑配置以及操作环境都不一样，使用的LORA以及插件版本、插件配置等也会有所区别，所以请不要追求一模一样的照片，创作您自己的完美作品吧！

4、如何取得联系？

如果有定制LORA的需求，或者有其他合作意向，可以直接在这里留言，或者联系Q331506796（注明来意）

5、你对差异提取法做了哪些优化？还是只是随便说说？

答：我对涉及代码进行了大量的优化：

1、自动化模型探索：自动从一组模型中生成多种融合模型，类似于参数调优或模型选择的自动化过程，但专注于模型融合。

2、基于欧氏距离的差异性分析：用于比较两个LoRA（Low-RankAdaptation）模型的权重差异。

在差异性分析中，欧氏距离用于量化两个模型权重之间的差异。具体来说：

权重表示：每个LoRA模型的权重可以看作是高维空间中的一个点。权重的每个元素对应空间中的一个维度。

计算差异：通过计算两个模型相应权重之间的欧氏距离，我们可以得到一个数值，表示这两个权重在高维空间中的相对位置差异。

解释差异：较大的欧氏距离意味着两个模型在该权重上的差异较大；较小的距离则意味着它们较为相似。

这种差异性分析用于指导模型融合的决策。例如，如果两个模型在某个权重上的差异很大，可能就需要特别处理这个权重，以确保融合后的模型能够有效地结合两个原始模型的特性。

这些优化完全是针对我自己的素材进行的，因为我的素材有一些特殊性，所以不是传统意义上的差异提取，要有一些取舍。

6、V3.0和V2.0到底有什么不同？

V3.0可以说是一个全新的LORA，可以说是全方位的改进。

一、拍摄

V2.0使用的是SK的极光。为了进一步提高质量，我购买了多款市面上比较好的产品进行对比，有极润K9、极润K1-2.0版、莱觅极光纯享2.0、莱觅极光、莱觅暗金2.0、凌琉光、JM妙不可言、SK极光高密版。

部分测试产品一览

最后经过测试，莱觅极光纯享2.0脱颖而出，效果最好。

莱觅极光纯享2.0

极润K9

极润K1-2.0版

凌·琉光

为了避免前两个版本不自然的情况，我对拍摄方法进行了特别的处理，比如加入了自然光下的素材，而不是都在影棚打光，这样出来的效果会更自然一些。（主要是要避免中间那一条线一样的光，否则出图会出问题，这也是这个版本重点要改善的地方。）

二、制作

制作中除了使用了自研的融合算法以外，还对数据集进行了深度处理，以确保效果。同时为了避免素材和实际效果不匹配的情况，我制作了两套LORA，智能化融合探索后形成了50个模型。

三、测试

排除掉其中的25个样本后，下图是剩下的25个样本的比对效果：

测试效果重点关注如下几个方面：

1、在权重为1时不开修脸，脸部不出现崩坏，尽量保持大模型原本的出图效果。

2、在不添加衣服的描述词时，对黑丝以外的任何画面元素影响最小。

3、鞋子、人物动作、影响最小，不出现画面整体色调变化等。

最后选择了17和24号模型进入下一轮测试环节。

4、测试不同底膜下的出图效果进行比对，并对权重进行探索,特别是加入了很多LORA一起用时的情况。

5、可以看到，24的效果更好一些，所以最终选择了24号。

6、和全网其他所有的同类型模型横向比对。

6、做一个模型倒贴一千块，买嗨丝就花了七百块，一地鸡毛，解散。

V2.0:

Hello everyone, the V2.0 version is here, mainly updated in the following aspects:

1. Reduce the pollution to the feet. When using this LORA, there was sometimes a problem that the shoes could not be displayed correctly. Now this problem has been fixed.

2. To increase the overall sophistication, the data set was changed to bf16+1024 real shots, and then the differences were extracted to achieve better results.

3. Further reduce pollution, less pollution to the composition of the overall picture, less likely to cause short limbs, etc., and more convenient to use.

4. The crotchless element has been added. Now using the trigger word Crotchless pantyhose has a certain probability of producing crotchless pantyhose.

But I also want to honestly state the shortcomings:

The main drawback I see with this version so far is that it’s not bright enough. Although the texture has indeed improved and it has the feeling of aurora, it is not bright enough. This may be due to the stockings themselves and the lighting used when taking photos. This problem will be mainly solved in the next version.

Overall, it is more convenient and easier to use, and it is strongly recommended to update.

大家好，V2.0版本来了，主要在以下几个方面更新：

1、减轻对脚部的污染，之前用这个LORA的时候有时会出现鞋子不能正确显示的问题，现在这个问题已经被修复。

2、增加整体的精细程度，数据集改为bf16+1024实拍，然后差异提取，效果更好。

3、进一步降低污染，对整体画面的构图污染更少，更不容易出现短肢等情况，使用起来更方便。

4、增加了开裆元素，现在用触发词Crotchless pantyhose有一定概率会出开裆连裤袜。

但是我也要诚恳的说出缺点：

这个版本目前我看到的缺点主要是不够亮。虽然质感确实有提高，也有极光的感觉，但是不够亮。这可能是因为丝袜本身的原因，以及拍照时打光的原因。这个问题会在下个版本中主要解决。

整体来说，还是更加方便和易用了，强烈建议更新。

V1.0:

Description:

This is a black silk LORA that can enhance the texture of black silk in the image.

Don't be fooled by the small volume of this LORA!

The features of this LORA are as follows:

1, It is carefully crafted using the method of difference extraction, so the volume is very small. Besides affecting the texture of silk stockings, it has little impact on other content in the image.

2, All are made using real-life photos taken with high-definition cameras to ensure realistic effects.

3, Because it has little effect on other content in the image, it's no problem even if the weight is large.

Why make another one when there are already so many black silk LORAs? Because I found:

1, Existing LORAs will pollute the image in some way. Either it's clothes, shoes, or can we make a LORA that only affects the texture of black silk and doesn't affect other content in the image?

2, The existing LORAs do not subdivide the thickness and style of silk stockings, and the overall state is quite chaotic.

So I made this LORA, and I will also make a series of different thickness, texture LORAs later. All are made with real shot materials, using the difference extraction method to ensure quality.

So this might be the best high silk LORA you've ever used.

Brand: LimerXXXX M

Series: Aurora

Thickness: 5D

Color: Black

Special attention! The weight of this LORA needs to be set between 1 and 2 to have a better effect. The weight needs to be set relatively large! Set other parameters according to habit.

Recommended LORAs to be used in combination:

tutu's face/图图的脸模/花ひよこ顔モデル

ADD DETAILS

OC

LOWRA

epiNoiseoffset

Film style

I usually post my works on TikTok.

My TikTok name is 花ひよこ, you can find it by searching directly. If you encounter any problems, you can leave me a message.

Looking forward to your works!

Looking forward to your works!

Looking forward to your works!

Disclaimer:

1, You should take full responsibility for any creative works you create using this model.

2, You can't use the model to deliberately create or share illegal or harmful output or content, avoid using the model for malicious, harmful, defamatory, fraudulent, or political purposes.

3, If you use this model for commercial purposes, please notify me, thank you.

Frequently asked questions:

1, Why can't I get a photo like you?

Everyone's computer configuration and operating environment are different, the LORA and plugin versions used, plugin configuration, etc. will also be different, so don't pursue exactly the same photo, create your own perfect work!

2, What to do if a large area of mosaic appears in the picture after increasing the weight?

Try to reduce the weight, or reduce the weight of other LORAs, or turn on the LoRA Block Weight plugin.

3, How to get in touch?

If you need to customize LORA, or have other intentions of cooperation, you can leave a message directly here, or contact via TikTok. Contact Q331506796 (state your purpose).

品牌：某觅

所属系列：极光

厚度：5D

颜色：black

简介：

这是一个黑丝LORA，可以改善图中黑丝的质感。

不要被这个LORA如此小的体积所迷惑！

本LORA的特点如下：

1、使用差异提取法精心制作，所以体积很小。除了会对丝袜的质感产生影响以外，对画面的其他内容影响较小。

2、全部采用真人照片高清相机拍摄制作，确保效果真实。

3、因为对画面其他内容影响较小，即使权重开的很大也没有关系。

为什么已经有了这么多黑丝LORA，还要再做一个？因为我发现：

1、现有的LORA都会对画面产生一些污染。要不然是衣服，要不然是鞋，可否做一个LORA，只会对黑丝的质感产生影响，而不会对其他画面内容产生影响？

2、现有的LORA都没有对丝袜的厚度、样式等进行细分，整体是一个比较混乱的状态。

所以我制作了这个LORA，后期还会制作一系列不同的厚度、质感的LORA，均采用真实实拍素材，使用差异提取法进行制作，确保质量。

所以这有可能是你用过的最好的嗨丝LORA。

特别注意！这个LORA的权重要在1到2之间才会有更好的效果，权重需要开的比较大！

其他参数根据习惯进行设置即可。

配合使用的LORA推荐：

tutu's face/图图的脸模/花ひよこ顔モデル

ADD DETAILS

OC

LOWRA

epiNoiseoffset

胶片风

如果您在使用中遇到问题，可以给我留言。

期待您的作品！

期待您的作品！

期待您的作品！

声明：

1、您应对使用此模型的任何创意作品承担全部责任。

2、您不能使用该模型故意制作或分享非法或有害的输出或内容，避免将此模型用于恶意、伤害、诽谤、诈骗或政治用途。

3、如果将本模型用于商业用途，请通知我，谢谢。

常见问题：

1、为什么我出不了和你一样的照片？

每个人的电脑配置以及操作环境都不一样，使用的LORA以及插件版本、插件配置等也会有所区别，所以请不要追求一模一样的照片，创作您自己的完美作品吧！

2、权重调高以后图片出现大面积马赛克怎么办？

尝试调低权重，或者降低其他LORA的权重，或者打开LoRA Block Weight插件。

3、如何取得联系？

如果有定制LORA的需求，或者有其他合作意向，可以直接在这里留言，或者联系Q331506796（注明来意）

概要：

これは、ブラックシルクのLORAで、画像の中のブラックシルクのテクスチャを改善することができます。

このLORAの小さな体積に惑わされないでください！本LORAの特徴は以下の通りです：

1、差異抽出法を使用して丁寧に作成したため、体積は非常に小さい。シルクストッキングのテクスチャに影響を与えるだけでなく、他の画像コンテンツにはほとんど影響を与えません。

2、全ての写真はハイクオリティのカメラで撮影した実物を使用し、リアルな効果を確保しています。

3、他の画像コンテンツにほとんど影響を与えないため、重量を大きくしても問題ありません。

なぜ既に多くのブラックシルクのLORAがありながら、もう一つ作る必要があるのでしょうか？それは以下の理由からです：

1、既存のLORAは全て画像に何らかの汚染を与えます。衣服であったり、靴であったり、ブラックシルクのテクスチャだけに影響を与え、他の画像コンテンツに影響を与えないLORAは存在しますか？

2、既存のLORAはシルクストッキングの厚さやスタイルなどを細分化しておらず、全体的に混沌とした状態です。

それゆえに、私はこのLORAを作成し、後期にはさまざまな厚さ、質感のLORAを作成する予定です。全てが実際のショット素材を使用し、差異抽出法を使用して作成し、品質を保証します。

したがって、これはおそらくあなたが使った中で最高のハイシルクのLORAになるでしょう。

ブランド：某觅

シリーズ：オーロラ

厚さ：5D

色：black

特に注意！このLORAの重さは1から2の間��設定すると、より良い効果が得られます。重量は比較的大きく設定する必要があります！他のパラメーターは、習慣に従って設定します。

推奨されるLORAの組み合わせ：

tutu's face/图图的脸模/花ひよこ顔モデル

ADD DETAILS

OC

LOWRA

epiNoiseoffset

フィルム風

私は普段、TikTokで作品を公開しています。

私のTikTokの名前は花ひよこです、直接検索すれば見つけられます。もし問題があれば、私にメッセージを送ってください。

あなたの作品を楽しみにしています！

あなたの作品を楽しみにしています！

あなたの作品を楽しみにしています！

免責事項：

1、あなたはこのモデルを使用して作成した全ての創作物に対する全責任を負います。

2、あなたはこのモデルを使って、故意に違法または有害な出力やコンテンツを作成または共有することはできません。このモデルを悪意を持って、傷つける、誹謗する、詐欺する、政治的な目的で使用することは避けてください。

3、このモデルを商用目的で使用する場合は、私に通知してください。ありがとう。

よくある質問：

1、なぜ私はあなたと同じような写真を出すことができないのですか？

各人のコンピューター設定や操作環境は異なり、使用するLORAやプラグインのバージョン、プラグインの設定なども異なるため、全く同じ写真を追求するのではなく、自分だけの完璧な作品を作成しましょう！

2、重量を上げた後に画像に大面積のモザイクが現れたらどうすればいいですか？

重量を下げるか、他のLORAの重量を下げるか、LoRA Block Weightプラグインを開くことを試みてみてください。

3、どのように連絡を取るべきですか？

LORAのカスタマイズが必要な場合、または他の協力の意向がある場合は、直接ここにメッセージを残すか、TikTokでメッセージを送ることができます。Q331506796（目的を明記）で連絡を取ることも可能です。

Files with SHA256: d1be2ee4508d7c0ac6ad30e6221f4eed236d1b81b05517a9bfcc906632d93e25