Type | |
Stats | 414 |
Reviews | (68) |
Published | Sep 24, 2023 |
Base Model | |
Training | Steps: 10,000 Epochs: 10 |
Usage Tips | Clip Skip: 2 |
Hash | AutoV2 DCCD152074 |
制作这个模型的目的是想借鉴mj中出的图片,具有很不错的光影质感和比较丰富的灰度层次,以让画面呈现所谓的电影感,让暗部的细节更丰富,亮部的曝光更柔和,同时希望增加一些光线粒子的感觉,让画面的真实性更有说服力。当然这个版本这方面做的还不够,相信随着版本更迭,会逐渐补足这个点。
模型的泛化性很好,兼容的艺术风格也比较多,可以和各种lora结合使用。没有什么必须的触发词,大家都可以尝试。采样器的话,以作者目前的尝试来看,并没有特别不适应的采样器,如果大家有遇到,可以浏览反馈。
推荐画布尺寸调大一些,生成的效果会更好。
_____________
The purpose of creating this model is to draw inspiration from the images in mj, which have a good light and shadow texture and rich grayscale levels, in order to present the so-called cinematic feeling in the image, enrich the details of the dark parts, and soften the exposure of the bright parts. At the same time, we hope to add some sense of light particles to make the authenticity of the image more convincing. Of course, this version has not done enough in this regard, and I believe that as the versions change, it will gradually make up for this.
The model has good generalization and is compatible with various artistic styles, which can be combined with various LORAs. There are no necessary trigger words, everyone can try. As for the sampler, based on the author's current attempts, there are no particularly unsuitable samplers. If you encounter any, you can browse the feedback.
It is recommended to increase the size of the canvas to produce better results.