Anime Illust Diffusion

AIDV2.10 是一次重大更新。它使用 两倍 于 aidv2.91 数据量的数据集大小和更高的图像品质训练得到。训练图像的分辨率由原来的768像素提高到了 1024 像素，这意味着您能够直接生成1024像素（比例可变，如768x1532）的图像而不会遭受图像畸变。同时，它有着更为平衡的风格权重。另外，aidv2.10 支持 200 余种不同插画风格（aidv2.91 为 100 种左右，较原先增加约 100 种）。同样地，AIDV2.10 有着其专属负面文本嵌入 aid210，即封面图中的 bad17。没有它，您很可能会生成出很丑陋的图像。您可以将它放在负面提示词的第一个位置以达到预期的效果。模型封面图均由纯文生图，即不使用任何Lora和ControlNet生成，部分经过高分辨率修复或二次图生图（以完全相同的参数）放大。aidv2.10 is a heavy update. It is trained with twice the dataset size of aidv2.91 and higher image quality. The resolution of the training images has been increased from 768(x768) pixels to 1024(x1024) pixels, which means that you can directly generate images of 1024 pixels (or variable scale ratio, such as 768x1532 and 512x2048) without suffering image distortion. At the same time, it has a more balanced style weight. In addition, aidv2.10 supports more than 200 different illustration styles (aidv2.91 is about 100).Likewise, AIDV2.10 has its own negative text embedding aid210, which is 'bad17' in the cover image. Without it, you're likely to generate ugly results. It's a good choice to utilize it and you may put it at the first place of your negative prompt.The cover images of the model are all generated by text2image, that is, generated without using any Lora or ControlNet. Some of them have been postprocessed with high resolution fix or secondary image2image (with exactly the same parameters as t2i).

animeIllustDiffusion_aidv210.safetensors

aidv2.10

aidv2.10_beta1

AIDV2.91 - Weak

AIDV2.91 - Strong

AIDV2.9

AIDV2.8

AIDV2.7

AIDV2.6

AnimeIllustDiffusionV2.5

AnimeIllustDiffusionV2.0

AnimeIllustDiffusionV1.0

Files with SHA256: 4efd47fad95d7dae0e2604a0f01d8c7fcaf0ee610ae3bd7f79a7447cbca17195