The model training set has a large number of steps, and its versatility is quite good. We hope everyone can download and generate it to support us, so we can save some computing power and electricity costs. Thank you!
Model Description:
This is a Lora for Internet commerce, UI illustrations, graphic design, commercial picture books, etc. The style is minimalist and geometric. The training set covers a very wide range of topics, and I tried my best to provide samples in various aspects such as people, animals, plants, architecture, environment, seasonal weather, and vehicles, in order to pursue better and wider generalization.
.
Recommended graphic size: The minimum single side of the image should be >=1024.
Trigger words: A picture in the style of CUI_3Dgss. After testing, there isn't much difference whether to use it or not. If the elements of the original picture are unpopular, you can add it if it doesn't follow the rules.
Sampling algorithm (Sampler): Euler a
Scheduler: simple, normal, or beta.
Sample frequency: About 30 steps. Some elements in the picture are broken and unclear, so we need to increase the number of steps.
Model weight: 0.8. Different weights will have different effects.
CFG: 3.5
CLIP: If the limb damage is severe, adjust the clip, 1 or 2; the clip parameters are not fixed.
High-definition restoration: Enlargement algorithm recommendation: R-ESRGAN_4x+Anime6B, 4x-AnimeSharp.
Redrawing amplitude: 0.3 (The smaller the value, the more it looks like the original image; the larger the value, the more it allows you to use your imagination.)
.
模型训练集大步数多，泛用性较好，希望大家下载生成支持一下，少亏一些算力和电费谢谢
模型说明：
这是一款用于互联网商业、UI配图、平面设计、商业绘本等的Lora，风格为简约几何风格。训练集题材涵盖非常广，我尽可能的在人物、动物、植物、建筑、环境、季节天气、载具等各个方面都提供了样本以追求更好更宽的泛化。
推荐图形尺寸：图片最小单边>=1024
触发词：A picture in the style of CUI_3Dgss，测试下来用不用差别不大，若生图元素小众，不遵从可添加试试。
采样算法（Sampler）：Euler a
调度器：simple，normal或beta
采样次数：30步左右，画面有些元素崩坏不清晰提高步数
模型权重：0.8，不同的权重有不一样的效果
CFG：3.5
CLIP：如果肢体崩坏严重就调clip，1或2；clip参数不是固定的
高清修复：放大算法推荐：R-ESRGAN_4x+Anime6B,4x-AnimeSharp,
重绘幅度：0.3（越小越像原图，越大越发挥想象）
.
Reference example of prompt words:
We recommend describing the keywords in natural language and polishing them with a large language model. Also, make sure there are no unclear references, as AI might interpret them differently.
提示词一：four people are dancing on a gray square floor, with a dark gray carpet in the middle, these people are wearing brightly colored retro-style clothes: a man in a green shirt and yellow pants, a woman in a blue top and purple skirt, another man in an orange shirt and green pants, and a woman in a red dress and black shoes, they are all excitedly raising their hands, there is a blue speaker on the right, and black star-like sparkles surround the dancers in the center, the background is pure white
提示词二：a man with dark hair, wearing glasses, a blue long-sleeved shirt, light blue pants, and black shoes, is standing on a green map, which has beige roads on it, he is holding a smartphone in his right hand and seems to be checking directions, there is a red location marker on the left side of the map, in the background, there are two translucent blue clouds floating, using flat color, with very few shadows
.
[Disclaimer]
1. In principle, this model is only for communication and learning, and the model and generated images cannot be used for commercial purposes.
2. The user shall handle any intellectual property disputes arising from the content generated based on this model on their own, provided that they comply with local laws and regulations and the terms of service.
3. Without the author's permission, you cannot integrate this model into any third-party software or services on your own.