Type | |
Stats | 606 2,947 |
Reviews | (95) |
Published | Jul 28, 2023 |
Base Model | |
Training | Steps: 1,764 Epochs: 10 |
Usage Tips | Clip Skip: 2 Strength: 0.9 |
Trigger Words | frilled dress, hair ribbon, dark skin, hair flower, black hair, long hair, red dress, kate, (shadow:1.2), dark-skinned female |
Hash | AutoV2 600CC8A7A2 |
Changes in version 1.0
improved training
added more images to the dataset
although it was trained with screenshots of the anime it is quite flexible with style loras and many models.
remove the clothing tags to use other clothes.
Problems: sometimes it will generate the face or just the eyes, add the "(shadow:1.2)" tag to avoid this (it won't always do it, usually on the second try it will get it right). or you can also use the weight in 1, this will avoid the generation of face and eyes in most generations, don't forget to use the tag "(shadow:1.2)".
That would be all, enjoy it!
⚠️⚠️ Warning: this lora is a beta, so it may contain some errors when generating images ⚠️⚠️.
New lora from kate from shadows house, this lora was trained with 50 images using the anything v5 model, the weight I recommend is 0.7 to 1.
Things that I noticed when generating the example images sometimes the eyes come out, my solution is to place in the positive prompt ((no eyes)), the same way with the mouth ((no mouth)) and the hands ((black hands)) or (black gloves), like when generating images of the body without much clothing, (or without anything) sometimes it will generate it completely in black (as it should be), but other times (mostly) it will come out with normal skin color.
add "black silhouette" and "silhouette" to your prompts for better results
please leave me your comments and generations.
without more to write, enjoy it!