Model should work on base or edit. Strength around 0.8 plus/minus works great.
Follow the trigger framework/words as below, it is a bit sensitive to it:
For the armbinder:
"woman is wearing a [color] leather armbinder" works well most of the times, but
on QWEN EDIT if there's a forward facing character you may have to use "woman has her arms tied behind her back in a [color] leather armbinder"
After the armbinder prompt:
specify the color of straps and chest or shoulder style.
Chest will usually give the X-straps in front of chest
shoulder will go straight over the shoulders.
As in: "... with [color] [chest/shoulder] straps"
Pose is trained to show three quarter view to make the armbinder visible, and it can do rear view turned head too; trained prompts:
"Side view"
"Rear view" / "Three quarter turned around pose"
"Kneeling on floor" / bed / elsewhere
Collar prompt - trained to do thicker collars and/or posture collar
"wearing [color] leather neck posture collar"
" wearing [color] leather neck collar"
Outfit (IMPORTANT)
It tends to default towards nudity; so if you want outfits or working with edit characters you will have to prompt it in natural language to do what you want with the outfit
Putting it all together, examples:
a woman's arms are tied behind her back wearing a white leather armbinder with white chest straps, she's wearing a black leather pencil dress, standing in a field of hay with bright skies, eyes looking at viewer, wearing black leather neck posture collar
woman wearing a black leather armbinder with black shoulder straps, standing leaning to the side a little bit, wearing a red latex dress
the woman's arms are tied behind her back wearing a red leather armbinder with red chest straps, she's wearing a [outfit], standing at [location], rear view, three quarter turned around pose, eyes looking at viewer, wearing black leather neck posture collar
I am still testing the lora for consistency, so any feedback is always appreciated.
Description
Model should work on base or edit. Strength around 0.8 plus/minus works great.
Follow the trigger framework/words as below, it is a bit sensitive to it:
For the armbinder:
"woman is wearing a [color] leather armbinder" works well most of the times, but
on QWEN EDIT if there's a forward facing character you may have to use "woman has her arms tied behind her back in a [color] leather armbinder"
After the armbinder prompt:
specify the color of straps and chest or shoulder style.
Chest will usually give the X-straps in front of chest
shoulder will go straight over the shoulders.
As in: "... with [color] [chest/shoulder] straps"
Pose is trained to show three quarter view to make the armbinder visible, and it can do rear view turned head too; trained prompts:
"Side view"
"Rear view" / "Three quarter turned around pose"
"Kneeling on floor" / bed / elsewhere
Collar prompt - trained to do thicker collars and/or posture collar
"wearing [color] leather neck posture collar"
" wearing [color] leather neck collar"
Outfit (IMPORTANT)
It tends to default towards nudity; so if you want outfits or working with edit characters you will have to prompt it in natural language to do what you want with the outfit
Putting it all together, examples:
a woman's arms are tied behind her back wearing a white leather armbinder with white chest straps, she's wearing a black leather pencil dress, standing in a field of hay with bright skies, eyes looking at viewer, wearing black leather neck posture collar
woman wearing a black leather armbinder with black shoulder straps, standing leaning to the side a little bit, wearing a red latex dress
the woman's arms are tied behind her back wearing a red leather armbinder with red chest straps, she's wearing a [outfit], standing at [location], rear view, three quarter turned around pose, eyes looking at viewer, wearing black leather neck posture collar
I am still testing the lora for consistency, so any feedback is always appreciated.