Clam

Name: Clam
Rating: 5 (547 reviews)
Author: Johnhe

572

2.8k

Updated: Oct 5, 2024

character

anime

01281-2637936798-1girl, purple eyes, blush, long hair, twintails, solo focus, brown hair, bow, bangs, hair between eyes, blue bow, nose blush,.png

01294-1650069941-1girl, purple eyes, blush, long hair, twintails, solo focus, brown hair, bow, bangs, hair between eyes, blue bow, nose blush,.png

01391-2061455677-1girl, solo focus, , open mouth, smile, blurry, depth of field, upper body, green eyes, standing, official , outdoors, , w.png

01416-4214692529-1girl,solo focus, full body, looking at viewer, facing viewer, , sleeveless dress, casual, white dress,floral background, o.png

01414-3523984531-1girl,solo focus, full body, looking at viewer, facing viewer, , sleeveless dress, casual, white dress,floral background, o.png

01408-3602718828-1girl,solo focus, full body, looking at viewer, facing viewer, , sleeveless dress, casual, white dress,floral background, o.png

01413-2892542903-1girl,solo focus, full body, looking at viewer, facing viewer, , sleeveless dress, casual, white dress,floral background, o.png

Images hidden due to mature content settings

Download (3.97 GB)

Verified: a year ago

PickleTensor

This asset is only available as a PickleTensor which is a deprecated and insecure format. We caution against using this asset until it can be converted to the modern SafeTensor format.

Details

Type	Checkpoint Merge
Stats	984
Reviews	Very Positive (145)
Published	May 27, 2023
Base Model	SD 1.5
Hash	AutoV2 978E740F81

1 File

default creator card background decoration

Johnhe

License:

CreativeML Open RAIL-M Addendum

Images hidden due to mature content settings

"Clam" is named from using modes' name.

It has better performance in txt2img when using positive prompts to describe what you want without any quality tags, and you can use EasyNegative as negative prompt.

Ver.1.1 has more details (like background,body...).

Ver.1.2 has more anime-styled hands (but it may generate strange figures).

无需添加正面质量标签，直接在正面标签里面输入想要的内容，使用EasyNegative或者其它嵌入式负面标签作为负面标签即可出较好的图。

建议使用vae，模型由于通过数次融合已经呈现出比较灰的颜色效果，vae修正可以得到比较好的结果。使用NovalAI官方原版vae（animevae.pt f458b5c6）或者mse-840000等都能呈现出比较不错的颜色效果。

本模型是融合模型得到的平涂风格大模型，CLAM来自最初合并的大模型的首字母或者其中关键字母。后续又不断融合了更多的模型，但是没有继续改名。该模型名称可以翻译为“蚌”。

该模型目的是在保留平涂风的情况下，增加更多有效的细节表现，并克服一些过于写实的表现，并旨在增加细节的稳定性。

1.2版本修复了手的细节，相较1.1更加动漫化，不过调参不好或者在潜变量放大的时候，有一定几率出现一些奇怪的身体姿势。基本很好保留了1.1版本的各种优点。不过有些缺点仍然会比较明显，比如配色没有那么好，用色偏平淡，有种雾蒙蒙的色彩感。光线表现太过丰富，但是又有点不太合理。

1.3版本修复了一些过分的细节感，较好地优化了配色和1.2版本奇怪的身体姿势，在光线表现上也趋于更好。手部的表现与1.2版本相当，不过在手被遮挡或者双手交叉或者位于图片边缘的时候仍然有奇怪的表现，容易和其他东西发生融合。不过一定程度上提高了出图质量，在不使用lora和controlnet的情况下，相比上个版本更容易在文生图的情况下生成违和感较低的图片。建议在高清修复前的生成图像像素长宽控制在1024以内，推荐512*768或者600*960，横图竖图都能出，效果都不错。

有好的意见或建议也欢迎反馈。会在以后版本好好努力的！

hf:https://huggingface.co/Johnhex/Clam

https://huggingface.co/Johnhex/Clam1.1

https://huggingface.co/Johnhex/Clam1.2

https://huggingface.co/Johnhex/Clam1.3