Models>

川口イッサ【Kawaguchi Issa】イチカちゃん Ichika-chan 人物 Lora 复印炼丹法 正负丹

No Images Found

Try Our Text To Image Demo

Quick Generate
Select a style, write what you want to see and click Generate
What you want in the image.
Clear
Random
* This model is not available for use and might generate different characters compared to the ones seen in the images.

Model Description by Creator

イチカちゃん的Lora系列使用了十四张来自川口イッサ老师推特的图片。采用了复印炼丹法。作者在原有的复印炼丹基础上做出了一些调整。因此,即便训练集容量较小,在各个大模型上的出图也都能够较好的保留イチカちゃん的特征,并且拥有一定的服装和动作的泛用性。

当然,Lora的高质量,最重要的依然是川口イッサ老师老师在推特上发布的超高质量和完成度的原画。原画中身体结构完整生动,肢体和手指的细节呈现也很完善。为训练减少了非常多的不便。

如果对作者在上文提到的训练方法感兴趣,欢迎阅读作者在B站专栏撰写的相关文章。

链接:https://www.bilibili.com/read/cv26824519/

训练参数方面,在一阶段Lora与大模型的融合之后,二阶段训练过程中作者一共将14张图进行了400 repeats per picture per epoch的训练,并总共训练了10个epoch。每个epoch在泛用性方面的表现略有差异。为了方便选择,作者会把在推荐参数0.5-0.7内各个Lora的出图表现用4张xyz图片加以展示,并上传这10个epoch对应的Lora。可以选择自己喜欢的图片对应的Lora进行下载。

最后,希望大家能够去推特支持非常优秀的原图作者川口イッサ老师,以及

生成的图片请勿用于任何商业用途!!

The Lora series of Ichika-Chan uses fantastic images from Kawaguchi Issa's Twitter. I have made some adjustment of the trainning method this time. As a result, even with a smaller training set capacity, the outtakes on some famous Checkpoints are able to retain the characteristics of イチカちゃん relatively well and have good performance on generalization of different clothing and movement.

Of course, the most important factor in Lora's high quality is the superb quality of the original drawings that Kawaguchi Issa tweeted. The original drawings have a complete and vivid body structure, and the details of the limbs and fingers are well rendered. It reduces the inconvenience of training very much.

If you are interested in the training methods mentioned by the author above, please feel free to read the related articles written by the author Bilibili.

Link: https://www.bilibili.com/read/cv26824519/

In terms of training parameters, after merging Lora and the Checkpoint in the first stage, the author trained a total of 14 pictures with 400 repeats per picture per epoch and a total of 10 epochs in the second stage of training. each epoch's performance in terms of generalizability is slightly different. In order to facilitate the selection, the authors will show the performance of each Lora within the recommended parameters of 0.5-0.7 in 4 xyz images, and upload the Lora corresponding to these 10 epochs, so that you can download the Lora corresponding to your favorite image.

Lastly, hope you can check out how good images the original author Kawaguchi Issa has posted on Twitter, and

Please do not use the generated images for any commercial purposes!!

Images Generated With This Model

Loading...

Similar Models

原神 男生组   特别系列 Genshin Impact boys' special series
N/A
V2
原神 男生组 特别系列 Genshin Impact boys' special series
(0.00)
0
0 Ratings
clover day's school uniform
V1
clover day's school uniform
(0.00)
0
0 Ratings
四禧丸子-梨安常服Lian(SiXiWanZi) LoRA
四禧丸子-梨安常服Lian(SiXiWanZi) LoRA
(0.00)
0
0 Ratings
原神 男生组   特别系列 Genshin Impact boys' special series
N/A
V2
原神 男生组 特别系列 Genshin Impact boys' special series
(0.00)
0
0 Ratings
【Art Style】赤倉 | Akakura
【Art Style】赤倉 | Akakura
(5.00)
0
1 Ratings
四禧丸子-恬豆常服Tiandou(SiXiWanZi) LoRA
四禧丸子-恬豆常服Tiandou(SiXiWanZi) LoRA
(0.00)
0
0 Ratings
四禧丸子-梨安常服Lian(SiXiWanZi) LoRA
四禧丸子-梨安常服Lian(SiXiWanZi) LoRA
(0.00)
0
0 Ratings
四禧丸子-恬豆常服Tiandou(SiXiWanZi) LoRA
四禧丸子-恬豆常服Tiandou(SiXiWanZi) LoRA
(0.00)
0
0 Ratings
Bowser
V1
Bowser
(0.00)
0
0 Ratings
niji3-lora
V1
niji3-lora
(5.00)
0
1 Ratings

Train Models

Train AI Models On Your Own Pictures

You can use your own pictures to train AI models. Reimagine yourself in new ways with our AI-powered image generator.
Create pictures of yourself around the world, in your dream outfit, as a video game character, or in countless art styles. The possibilities are endless.
Train Your Own Model

Generate AI Images. Simple and Easy

AIEasyPic is the #1 AI image generator. Generate images with simple prompts, swap faces, train models to reimagine yourself, and more.