This is my second LoRA!
这是我制作的第二个Lora!
Made based on the character Mikage Sara from artist tahita's work The Great Detective is Captivated by the Investigation.
是基于艺术家tahita的作品《名探偵は捜査の虜》中的角色：御影サラ 所制作。
Base Model: waiIllustriousSDXL_v160
基础训练模型：waiIllustriousSDXL_v160
Recommended for use with anime-style base models. Best results are achieved with a weight around 0.8.
建议搭配动漫风格的基础模型使用，将权重调整在 0.8 上下可获得最佳效果。
采样器：Euler A
调度器：simple
步数：30-50
CFG：7
===============================================================
Trigger Word(激活词):mztyyy,
Base Prompt(基础):mztyyy, 1girl, solo, short hair, grey hair, brown eyes,
mztyyy，1个女孩，单人，短发，灰发，棕眼
Outfit 1(服装1):mzt brown hat, mzt brown pinafore dress, white collared shirt, black neck ribbon, mzt black gloves, mzt belt, black socks, mzt brown loafers,
MZT棕色帽子、MZT棕色背心连衣裙、白色有领衬衫、黑丝颈部丝带、MZT黑色手套、MZT腰带、黑袜子、MZT棕色乐福鞋
Outfit 2(服装2):mzt cat ear hairband, mzt red collar, mzt brown elbow gloves, mzt brown thighhighs, heart-shaped pupils, blush,
MZT猫耳发带，MZT红色项圈，MZT棕色肘部手套，MZT棕色过膝袜，心形瞳孔，脸红
Optional(可选):shoulder strap slipping, open clothes, breasts out, breasts, nipples,
肩带滑落，开衫，露出胸部，胸部，乳头white panties, pussy, cameltoe,
白色内裤，阴户，骆驼趾thinking, holding black magnifying glass,
思考，拿着黑色放大镜one eye open while the other is half-shut,
一只眼睁开一只眼半闭tongue out, clenched teeth, teeth, back view,
吐舌，咬牙，牙，后视图
Her outfits and prompt tags can all be split up and used individually or in combination.
她的所有服装或提示词都可以拆分并单独或组合使用。
The back designs of the character's outfits have also been completed and trained.
人物服装的背面也做了补全和训练。
There is an exclusive thinking pose!
有专属的思考姿势！
===============================================================
Final Notes：
I am not a professional LoRA creator. Making and publishing LoRAs is merely part of my learning process and a hobby.
Every creation and material completion is done during my holidays and after work hours.
Therefore, please do not expect high-quality or fast production from me.
The reference materials for this second character are extremely limited. I had to find usable materials from the manga, cut them out one by one, and retouch them manually. There are far too many texts and obstructing objects in the manga, and since I am not professional, many of my processing methods are manual or rather old-fashioned, which makes the work extremely time-consuming.
For instance, the back design of the character's pinafore dress straps can only be roughly seen as crossed on Page 4 of the first volume of the manga. I had to draw it manually based on imagination and train repeatedly until it blended naturally with the character.
After another seven or eight rounds of training and refinement, I finally achieved this relatively satisfactory result.
In addition, strong reflections sometimes appear on the character's body and hat. I tried adjusting parameters such as color, brightness and exposure of the training images, and even manually modified the colors of the reflective areas in Photoshop. After a long period of trial and error and training experiments, I still failed to solve this problem.
Is there any professional who can give me some guidance on this issue?
写在最后：
我并不是一个专业的Lora制作者，制作并发布仅仅是我的学习过程和爱好。
每一次的制作和素材补全，都是利用的假期和下班时间。
因此，不要奢望我能制作的多好或者多快。
这第二个角色的素材实在是有限，我需要尽量在漫画中找到可以利用的素材，将它们从漫画中一张一张的扣取并补绘。漫画中的文字和遮挡物体实在太多了，再加上我并不专业，很多处理办法都是手动或比较古早的方法，所以非常耗时。就比如角色背心连衣裙的背带的背身形态，漫画中仅仅漫画第一部的P4可以大致看到是交叉的形态的。我只能凭想象手工绘制，并反复训练至和角色没有违和感的融合。
又经过了七八轮的训练和改进，才得到了这个相对满意的结果。
另外角色身体和帽子上有时会出现很强的反光，我尝试调整训练素材图中的颜色亮度曝光度等参数，甚至手工用PS调整了反光区域的色彩。经过长时间的折腾和训练试验，结果还是失败了。有专业的人士可以指导一下这个问题吗？