Type | |
Stats | 2,300 |
Reviews | (342) |
Published | Jul 30, 2023 |
Base Model | |
Training | Steps: 8,000 Epochs: 15 |
Usage Tips | Clip Skip: 2 |
Hash | AutoV2 75052D03A1 |
To support my work, you could buy me a coffee
https://bmc.link/cyberdicklang
or 爱发电 赛博迪克朗
Yesterday, I made a new model to participate in the Civitai AI model competition. It can quickly generate fairy tale-style picture book illustrations. The light, shadow and details are flat, and elements such as stars and flowers that children like are added. It integrates the color expression of Pop Art. Very serious prompt words can be displayed in a dreamy and interesting form. Overall, I am quite satisfied. I made a bedtime story for my daughter, and she read it seriously
I have always been curious about how the LORA of this style of painting is overfitting, so I simply refined it so that it will not run away at all. After testing all kinds of modern scenes, they have basically achieved a better two-dimensional, and the colors are as fancy as ever.
You don't need trigger words, just install them and use them. During training, a lot of flowers and little stars were added, so if you want to achieve a better fairy tale effect, you can try these words.
“flower, star, magical”
The controllability of the prompt words may be improved in the future, and some different styles and types of picture books may be made. Too thrilling painting styles will not be considered.
If anyone can use this model to make some interesting children's stories, please share it with me, and I will show it to my children.
昨天做了个新模型参加C站AI模型大赛，可以快速生成童话风格的绘本插图，光影和细节都做到平面化，加入了小朋友喜欢的星星和花等元素，融合了波普艺术的色彩表现，可以把非常严肃的提示词以梦幻有趣的形式展现出来。整体来说还挺满意的 给闺女做了睡前小故事，看的可认真了
我一直好奇这种画风的LORA怎么算是过拟合，索性就给它炼到完全不会跑跑偏好了。测试了一下各种现代场景基本都达到了比较好的二维化，色彩一如既往的花里胡哨。
不需要触发词，装上就用。训练的时候加了很多花花和小星星进去，所以如果想达到更好的童话效果可以试试这几个词。
“flower, star, magical”
后续也许会改进提示词的可控性，大概可能会做一些不同风格类型的绘本风。过于惊悚的画风不会考虑。
如果有人能够用这个模型去做一些有趣的儿童故事，欢迎分享，我会给我家孩子看的。
Contact:
B站小红书 赛博迪克朗 欢迎点个关注吼
Email: [email protected]