Anime Illust Diffusion

与 AIDV2.6 相比，AIDV2.7 新增了7种风格（见附录A）。另外，AIDV2.7 添加了 0.05 的 offset_noise，生成的图像对比度更高。展示图中使用了特制的负面文本嵌入 aidv1（显示为badv12）。下载请参阅 <a target="_blank" rel="ugc" href="https://civitai.com/models/16807?modelVersionId=40158">https://civitai.com/models/16807?modelVersionId=40158</a>AIDV2.5 的构图融合了一些主流模型，人物姿势稳定；而 AIDV2.7 的构图虽然也融合了其他模型，但相比于 AIDV2.5 来说权重很低。这么做的缺点是构图不稳定；但优点是，AIDV2.7 学习了很多不寻常的构图，且拟合程度很高，有时能抽更惊艳的图像。我暂时不会放弃这种选择。我正致力于微调 AIDV2.7，让它能同时平衡构图和风格。Compared with AIDV2.6, AIDV2.7 adds 7 new styles (see Appendix A). Furthermore, AIDV2.7 introduced an offset noise of 0.05 resulting in higher contrast in generated images. The illustration uses a specially crafted negative text embedding called aidv1 (shown as badv12). For download see <a target="_blank" rel="ugc" href="https://civitai.com/models/16807?modelVersionId=40158">https://civitai.com/models/16807?modelVersionId=40158</a>The composition of AIDV2.5 incorporates some popular models, and the posture of the characters is stable; while the composition of AIDV2.7 also incorporates other models, but the weight is very low compared to AIDV2.5, and the degree of noise fitting is very high. The disadvantage of this is that the composition is unstable; but the advantage is that AIDV2.7 has learned a lot of 'unusual but special' compositions, and sometimes it can draw more stunning images. I won't give up this option just yet. I'm working on fine-tuning AIDV2.7 to balance composition and style.

animeIllustDiffusion_aidv27.safetensors

AIDV2.7

aidv2.10

aidv2.10_beta1

AIDV2.91 - Weak

AIDV2.91 - Strong

AIDV2.9

AIDV2.8

AIDV2.6

AnimeIllustDiffusionV2.5

AnimeIllustDiffusionV2.0

AnimeIllustDiffusionV1.0

Files with SHA256: 33fd6e75408914df494aa88a17bd64788f3080508d0e64becd6dd6ab0252590c