Sign In

下级生2D(下級生2D)

10
133
5
Updated: May 31, 2025
base model
Verified:
SafeTensor
Type
Checkpoint Merge
Stats
133
0
Reviews
Published
Apr 20, 2025
Base Model
SD 1.5
Usage Tips
Clip Skip: 2
Hash
AutoV2
5347B40B27
default creator card background decoration
HGOD's Avatar
HGOD

中文:

SD1.5的时代已经过去,我从收藏中的超5000个SD1.5模型中挑选了部分模型进行融合来进行发布,来纪念这SD1.5的光辉时光。

由于挑选模型含有大量限定模型以及未在civitai发布过的模型,所以请不要询问融合成分。

真人系模型已完成

-(同级生) 2025.1.31

-(下级生) 2025.2.2

-(上级生) 2025.2.3

2D模型

-(下级生2D) 2025.4.20

-(同级生2D) 2025.5.31

-(上级生2D) 2025.5.23

本次下级生使用3个限定模型以及1个LORA作为融合材料。

手脚相关是SD1.5的通病,可以自行使用手脚修复插件。

负提示词没有限制。embedding和TI插件也可以使用,不过推荐使用

badhandv4, (worst quality:2), (low quality:2), (normal quality:2), lowres, ((monochrome)), ((grayscale)), jpeg artifacts,signature, watermark, username, blurry, artist name, text, error,

推荐配置:

- Sampler: DPM++ 2M (series)

- Steps: 20-30 (recommended 30)

- CFG: 6-7 (recommended 7)

- 重绘幅度0.3-0.5 (recommended 0.5)

其它参数可以参见例图。

弱项:手脚相关需要使用修复手脚插件会有更好的效果,推荐使用高清修复以及Adetailer一起使用。相关参数可以参考示例图。年龄相关请自行解锁。

(10发岸宝,20发专武,谢谢岸宝!)因为封面上去就是LORA了,所以放后面让你当本次示例LORA女主

融合目标是良好的脸型+真实系的背影+良好的光效+黑白丝袜(雪糕控)。

只能说勉强达到,但是因为融合,导致各效果都跟原始模型效果有所降低,只能说有得有失吧。可能需要ROLL卡,因为打了细节加强LORA所以控图方面会欠缺,多ROLL卡总归会出一些还可以的。欢迎使用。

english

The era of SD1.5 has come to an end. To commemorate the glorious times of SD1.5, I have selected and fused several models from my collection of over 5,000 SD1.5 models for release.

Since the selected models include many exclusive models and models not published on Civitai, please do not inquire about the fusion components.

Realistic models have been completed:

-(同级生) 2025.1.31

-(下级生) 2025.2.2

-(上级生) 2025.2.3

2D Models:

-(下级生2D) 2025.4.20

-(同级生2D) 2025.5.31

-(上级生2D) 2025.5.23

The 下级生2D model this time uses 3 exclusive models and 1 LORA as fusion materials. Hand and foot-related issues are a common flaw of SD1.5; you may use hand/foot repair plugins as needed.

Negative Prompt Suggestions:

There are no limits on negative prompts. Embedding and TI plugins are also supported; however, the following are recommended:

badhandv4, (worst quality:2), (low quality:2), (normal quality:2), lowres, ((monochrome)), ((grayscale)), jpeg artifacts,signature, watermark, username, blurry, artist name, text, error,

Recommended Settings:

Sampler: DPM++ 2M (series)

Steps: 20–30 (recommended 30)

CFG Scale: 6–7 (recommended 7)

Strength: 0.3–0.5 (recommended 0.5)

Other parameters can be seen in example images.

Weak Points:

Hands/Feet: Using hand/foot repair plugins will give better results. It's recommended to use high-resolution restoration alongside Adetailer. Relevant parameters can be referred to from the example images.

The fusion target is a good face shape + realistic-style background + good lighting effects + black and white stockings (ice cream theme).

I can only say it barely achieves the desired results. However, due to the fusion process, the effects of each element are somewhat reduced compared to the original model's effects. It's a trade-off where you gain some and lose some. It might be necessary to roll some cards because adding the detail-enhancing LORA tends to compromise control over the image. Rolling more cards will eventually yield some satisfactory results. Feel free to give it a try!