Type | |
Stats | 50,544 |
Reviews | |
Published | Feb 23, 2023 |
Base Model | |
Hash | AutoV2 E7BF829CFF |
Night Sky YOZORA Model ——“For ultimate image quality and large image size(>1536 x 1024)”
Trained by @YozoRaAru
如果你需要更好的色彩表现,我推荐你试一下Color Box
If you want better colour representation There is a new style model Color Box
YOZORA is a model that pursues perfection and is filled with personal preferences. I trained it using images that I like, which give it an unparalleled level of detail and completion. It can provide you with very exquisite character images.
In addition to training, I also blended it with dozens of other high-quality models that I personally like, and I tried layer blending as well. Among these models, I selected NovelAI's original model as the major weight because it can give it excellent adaptability to a variety of prompts. I haven't tried testing it with landscape + character yet, but I believe it can also perform excellently in that aspect.
For the best results, please ensure that your final resolution > 1536x1024. This is necessary for generating exquisite illustrations
Here are some recommended settings for the parameters:
Clip skip: It is recommended to set it at no less than 2. YOZORA already has a rich level of detail, and setting it at 1 may make the image appear cluttered and confusing.
Resolution: YOZORA may not be suitable for generating small images because the extreme level of detail may become too crowded. It is recommended do not use Hires.fix to generate large images. A larger resolution is equivalent to a larger canvas, making it easier for YOZORA to capture details. It is recommended to use 1536 x 1024 or higher for Hires.fix.
Negative prompts: It is recommended to use EasyNegative to provide brief and precise descriptions.
The name YOZORA means "strolling among the stars in the night sky". I hope that she can bring you the excitement and joy of a sky full of stars on a starry night.