Ada-1 from Destiny 2.
-
The model was trained on a little over 50 images
-
I'm happy with the result of this model because this version is much more forgiving with the tags so you don't need to be as specific. Also there is a better consistency in the images generated, even across different checkpoints and styles.
-
To get better details in the face, try to generate at a higher resolution. Also using High-res fix helps when generating too.
-
sometimes using ADetailer for the face can give better results depending on the checkpoint and other loras used.
-
This will be the final version of this model for SD 1.5
Description
This is the final release of this model for SD 1.5.
Added more training images
optimized previous training images and captions
The tags have been simplified and the training settings have been optimized so that the model is significantly more consistent and flexible.
Works the best with realistic checkpoints, but still works well with more stylized ones.