I’m thrilled to present Evalisennia v3.0—the best SDXL model I’ve ever created! It took months of training, and I’m excited to finally share it with you. Please enjoy, share your results, and leave comments!

Evalisennia v3.0 is a photorealistic SDXL model designed to generate highly realistic, diverse East Asian female subjects. It excels at producing images that look convincingly like real photographs, with vivid colors, rich details, and outstanding prompt comprehension.

如果你喜欢这个模型，欢迎多多分享你的作品并与我交流，我会非常开心看到大家使用它所创作的图片，拜托了！

Key Features | 模型特点

Extreme Realism | 极度真实
High Definition & Rich Details | 高清晰度与细节
Great Prompt Interpretation | 强大的提示词理解力
Unique Faces, Not Repetitive AI Faces | 拥有独特面孔
High Flexibility with Diverse Training Data | 高强灵活性，丰富数据
- 虽然本模型主要用于逼真写实风格，如果在提示词里使用其他类型（如 3D、插画风格）也能获得不错的结果。
- 训练数据包含约 7000 张图片，涵盖了许多服饰和场景，只要提示词描述得当，模型会尽力还原！

Recommended Parameters | 推荐参数

Prompt:
high-resolution
Negative Prompt:
worst quality, low quality, illustration, 3d, 2d, painting, cartoons, sketch, bad hand, bad fingers, bad anatomy, average quality
Sampling Method | 采样方法:
- DPM++ SDE Karras (30+ steps)
- DPM++ 2M Karras (50+ steps, faster)
Image Size | 分辨率:
- 832 × 1216
ADetailer:
- For distant faces, consider using ADetailer (inpaint padding = 128).
- 建议在设置中开启 Dynamic Denoising，这样无论远景或近景都能让 ADetailer 自动调节，不必手动开关。

Credits

Huge thanks to the creators of the following great models used in the merge:

LEOSAM's HelloWorld XL by LEOSAM
RealVisXL V5.0 by SG_161222
XXMix_9realisticSDXL by Zyx_xx

Please enjoy generating with Evalisennia 伊岚 v3.0! I look forward to seeing your creations and hearing your feedback. Have fun!