Trigger Word has now been changed for better tuning with braided sides
+Works perfectly with inpainting.
Link to SD1.5 version made by Wolf_Systems
How to use,
Must important include "braided side" somewhere in the prompt.
You can still prompt other hairstyles combined with the braided sides.
Prompt Example: "Braided Side, Mohawk"
Although not trained on enough images it can still make the braided side in a different color from the rest of the hair. (Results may vary)
Prompt Example: "Pink braided side, Blonde hair,"
Important tip:
To combat profile perspective:
Positive prompt: "Looking at viewer"
Negative prompt: "profile"
Avoid using too short prompts as that will cause the model to have higher control of the perspective.
Known Issues: Braids may occasionally transform into non-braided hair. This wasn't an issue in the alpha version because it utilized two triggers: one for half braids and one for full braids. I might reintroduce this functionality in a future model update, but for now, you can always revert to using the alpha, although its results are generally inferior.
Description
First Non-Alpha
Trigger Word has now been changed for better tuning with braided sides
+Works perfectly with inpainting.
How to use,
Must important include "braided side" somewhere in the prompt.
You can still prompt other hairstyles combined with the braided sides.
Prompt Example: "Braided Side, Mohawk"
Although not trained on enough images it can still make the braided side in a different color from the rest of the hair. (Results may vary)
Prompt Example: "Pink braided side, Blonde hair,"
Important tip:
To combat profile perspective:
Positive prompt: "Looking at viewer"
Negative prompt: "profile"
Avoid using too short prompts as that will cause the model to have higher control of the perspective.
Known Issues: Braids may occasionally transform into non-braided hair. This wasn't an issue in the alpha version because it utilized two triggers: one for half braids and one for full braids. I might reintroduce this functionality in a future model update, but for now, you can always revert to using the alpha, although its results are generally inferior.