Introduction: Patch optimisation version, adjust the picture composition, optimisation of special effects, additional mechanical texture, real texture and comic effects, for the face is still largely biased towards the previous characters, but because of this increase in some of the European and American face did not lock the face, so such as Europe and the United States Department of the face can be used (ukl, black hair, bangs, Chinese style, and other words trigger the Chinese style), no counter words will be biased grey tone, counter words to increase the quality of words will make the picture colour contrast more strongly out of the picture will also be more convergent, specifically you can correspond to the way you like the picture. ), no anti-word will be partial grey tone, anti-word to increase the quality of the word, etc. will make the screen colour contrast is more intense out of the picture will be more convergence, specific you can correspond to their own preferred way out of the picture.

The end of the newbie recommended sentence: 1girl, white hair, cityscape, first try!

I hope you have fun, more use of the scene also hope that you explore more ah.

I'm not sure if you're going to be able to do this, but I'm sure you'll be able to!

v1048

-This model is a 2.5d realistic style model based on sd xl1.0.

-Recommended size: 768*1344, 1024*1024, 1024*1280 and so on.

-Recommended steps: from 30

-Recommended sampling: [Euler a] [DPM++ 3M SDE Exponential] [Restart].

-Recommended antonyms: a graph without antonyms is [None] (malformed hands, poorly drawn hands, mutated fingers: 1.4), -cfg scale: 3.4, -cfg scale: 3.5

-cfg scale: 3.5 to start with

-Updated 5 June 2024 I'll try to iterate on this.

-Zoom in recommendation: use Tucson to zoom in, sampling is the same as in the outgoing image, step is 1.5-2 times of the outgoing image, redraw amplitude 0.1-0.2

-Attention: the whole body open ad repair face Recommendation: [face_yolov8n_v2.pt], a large number of running map is recommended to open the ad, my high-definition map is basically Ultimate SD upscale zoom processing, zoom algorithm [4x-UltraSharp].

Casually introduce the 2 routes:

Lifelike XL is on the realistic side.

Fantasy Hacker XL on the special effects side.

---------- ---------- ---------- --------

Model Usage Statement:

1. Please do not host this model and its derivative versions (e.g. fusion model versions) on websites or applications intended for profit or fundraising. If you want to use this model and its derivative versions for commercial purposes (e.g. generative services, sale of images, posting of articles on short video platforms or use of images in publications, etc.), please contact the original author for permission.

2. Do not directly sell the images generated by this model without sufficient manual modification to make it an independent work.

3. Do not use the model to intentionally create or distribute illegal or ???????????????. Please ensure that the model is used for positive and active purposes and in accordance with legal and ethical guidelines. 4.

4. The model and the generated images are intended for personal entertainment, scientific research and hobby exchange only, and commercial use is prohibited.

The user is solely responsible for any legal consequences and the original authors are not liable.

This statement applies to all versions, and please credit the original author when using the fusion model. Thank you for your respect and understanding of the modellers.

Email: 470787985@qq.com

DanYaoSquare-AIGC Exchange Group:261454498

介绍：补丁优化版本，调整图片构图,特效优化，追加了一些机械质感，真人质感和漫画效果，对于脸型大体还是偏向于之前的人物，但是因为这次增加了一些欧美脸型没锁脸，所以如出现欧美系的脸型可用（ukl,black hair,bangs,Chinese style等词触发中式），无反词会偏灰色调，反词增加质量词等会让画面颜色对比更加强烈出图也会更加的收敛，具体大家可以自行对应喜欢的方式出图.

完新手推荐句：1girl,white hair,cityscape,先尝试

希望大家玩的开心，更多使用场景还望大家多多探索呀。

多多返图~！！

v1048

•本模型是一个基于sd xl1.0的2.5d偏写实风格类模型

•推荐尺寸：768*1344，1024*1024，1024*1280等

•推荐步数：30起

•推荐采样：【Euler a】【DPM++ 3M SDE Exponential】【Restart】

•反词推荐：没有反词的图就是【无】(malformed hands, poorly drawn hands, mutated fingers:1.4),

•cfg scale：3.5起步

•更新2024年6月5号我会努力迭代进步

•放大推荐：使用图生图进行放大处理，采样和出图时一致，步数是出图的1.5-2倍，重绘幅度0.1-0.2

•注意：全身开ad修脸推荐：【face_yolov8n_v2.pt】，大批量跑图建议开启ad，我的高清图基本都是Ultimate SD upscale放大处理，放大算法 [4x-UltraSharp]

随便介绍下 2条路线：

栩栩真境XL偏真实类

奇幻骇客XL偏特效类

——————————————————————————————————————

模型用途声明：

1. 请勿将此模型及其衍生版本（如融合模型版本）托管于旨在盈利或募捐的网站或应用程序。若需将此模型及其衍生版本用于商业目的（如生成式服务、图片销售、在短视频平台发布文章或出版物中使用图片等），请联系原作者授权。

2. 未经充分人工修改以使其成为独立作品的情况下，不得直接销售此模型生成的图片。

3. 请勿利用该模型故意制作或传播非法或有害内容。请确保将模型用于正面积极用途，遵守法律和道德准则。

4. 此模型及生成图片仅供个人娱乐、科研和兴趣交流使用，禁止商业用途。

对于任何法律后果，使用者将自行承担责任，原作者不承担任何责任。

此声明适用于所有版本，并请在使用融合模型时注明原作者。感谢您对模型制作者的尊重和理解。

邮箱：470787985@qq.com

丹耀四方-AIGC交流群:261454498

白棱XL_奇幻骇客 bailing XL_Fantastic hacker