An experimental attempt to test the boundaries of Krea. It does most things quite well, but at the cost of some minor breaking here and there because mixing of concepts. Not as perfect as a specialized lora for each concept, but good for general creation. Would appreciate if you test and provide feedback!
See example prompts in images. trained on 7000 steps around 450 images. All images posted with 8-12 steps, Krea Turbo, CFG 1, Strength 1.0 to 1.2. You can push strength up, usually doesn't worsen quality.
Krea is incredibly good with prompt adherence, so a lot of things can be fixed with prompt edits.
Concepts it knows, and this is how it was trained
Bondage Poses
General full body rope bondage (standing / sitting)
Chair Tied
seated bondage scenario, Woman is sitting on a [type of] chair bound with [color] [material:rope/tape]. [Add general description of how she is tied]
Hogtied
Woman is lying on a [surface] bound in a hogtied pose with her arms pulled behind her back and her legs bent upwards. She is bound with [color] rope, with the ropes tying her [wrists/elbows/arms/ankles/thighs/knees/chest/shoulders]. Her ankles are tied to her bound wrists."
Elbow Bondage
The woman is in a elbow bondage pose with her arms tied behind her back. White ropes wrap around her wrists and elbows. (Usually prompting for profile and/or three-quarter views work best)
Breast Bondage
A woman with [small/medium/large] sized breasts in a breast bondage scenario, [color] ropes wrapping above and below breasts, with additional ropes intersecting over her chest running behind her neck. [color] ropes are wrapped constricting the base of each breast.
Optional: Silver and black clamp clips are pinching her nipples with grey spherical weights hanging from the clamps. OR silver clamps are pinching her nipples connecting with a thin metal chain.
Strappado
A woman in a strappado pose, with her arms bound securely behind her back and elevated upward behind her torso. Her wrists and elbows are bound together behind her back with [color] [materials], and the binding is pulled upward, anchored vertically off-screen... [optional: ...specify if a different material is pulling her wrists up].
Spread Eagle
A woman is tied in a spreadeagle pose with her arms and legs outstretched. She is [standing/lying on her back/lying prone]. Her wrists and ankles are wrapped with [restraint: cuff/ropes/etc.] anchored [environment location + anchor material like chains etc.].
Arms Over Head
A woman in a static column tie pose, she has her wrists bound with her arms secured extended fully above her head. Her wrists are bound together with [color] [material:ropes/cuffs] anchored vertically to [point] with [material].
Mummified (tape or other materials)
The woman is [pose/position]. She is in a mummified bondage scenario, with all parts of her body cocooned under [color] [material:tape/leather] encasing her continuously from [shoulders/neck] to [waist/thighs/ankles/toes], [optional:with a cutout for her breasts].
Gags (See Issues Note Below)
Ball Gag: The woman has a [color] ball gag in her mouth fastened with [color] leather straps.
Cleave Gag: The woman has a strip of [color] [material: cloth, rope etc.] cleave gag between her lips fastened behind her head.
Wraparound Tape Gag: The woman has [color] tape gag wrapped tightly around her mouth looping behind her head.
OTM: The woman has a wide strip of [color] cloth covering her mouth as a gag fastened behind her head [add with a knot if needed]
Ring Gag: The woman has a [color] ring gag [in her mouth/between her teeth] fastened with [color] leather straps.
Panel Gag: The woman has a [color] leather panel gag over her mouth fastened with black leather straps.
HARNESS STRAPS: Change the "leather straps" in the ball/ring/panel above to "leather face harness straps".
NOTES:
Ring gag works good in close ups but will definitely struggle in wider/zoomed out perspectives.
Harness straps can also be a miss in its shapes a lot of time. If i change the prompt to add "triangular leather face harness straps bridging over her nose" it usually fixes it but still can be a miss in zoomed out perspectives.
Description
FAQ
Comments (23)
GREAT WORK!!
Thank you!
😍 👍🏻
work amazing with other character loras
That’s great to know thank you for the feedback, I hadn’t tested it
Works well!
Amazing work thank you ^^!
Thanks for the feedback! Much appreciated
Works much better than I expected. Fantastic job and thanks for sharing this gem!
Thank you for the feedback :)
Awesome Lora! Would love to see plug gag and spider gag make an appearance in a future version!
I will keep that in mind! Thanks for the feedback
I love all your LoRAs, but with this one I am struggling with degraded image quality. No matter which resolution or which LoRA strenght I try, the output looks blurry and almost overfitted (as if the strength was turned up to high). But I am running strength between 0.5 and 1.0. I use Krea 2 Turbo with 8 steps. Any idea?
How are your images' quality compared relatively to my samples or all the other samples posted in the gallery for the LoRA? Better or worse? If they're worse on average then its probably not the LoRA but the workflow used or the VAE used because Krea has those frizzy hair tendencies if the prompt is not defined enough for quality, lighting, etc.
@vackyadmiral167 Thank you for taking the time to respond. Your samples look great, mine are way worse in terms of quality. I guess I have to work on my prompting then? I didn't give many details regarding quality or lightning i guess.
@dachhaus41200 Hey no worries. For sure try promoting but that alone won’t fix it - see below
First off change your steps to 10 or 12. Even though Krea turbo recommends 8, 10-12 is the default where people are getting good results.
If that doesn’t fix it, try looking the workflow, checkpoints, or VAE etc you’re using - have a look at the ones posted on CivitAI and/or some articles around because there’s a lot of know issues and people have issued great updated VAEs and checkpoints that will help better the quality even more.
@vackyadmiral167 Thank you very much, I will have a close look at each of the settings I am using in the workflow.
In the bottom left of each image I try to make with this lora, I'm getting some unwanted text. Usually it's text of part of the prompt I am using. I'm not sure how to get rid of that text?
Yeah because of the logos of the images it was trained on. Shitty side effect. Just prompt away "Do not produce any watermark logos or text in the image" or something of the sorts. usually works for me.
Great stuff, as you said some pises work better than others but can be tuned with some careful prompting.
Would you consider making the opposite version where the male is in the bondage role? I tried prompting for role change but I think it leans very heavily towards female due to dataset.
I do love a nice friendly dominatrix scene😋
Hey, thanks for your feedback and for trying the LoRA!
I am afraid the solution might be simpler! I am using the Realism Engine lora to do gender swap for roles when i need to - works perfectly for me around strength 0.8 to 1.0 combined with my LoRA. Try this out and see if it works for ya!
https://civitai.red/models/2688234/realism-engine-ideogram-4-krea-2
@vackyadmiral167 ok, thanks for the tip.i was using a different safety bypass
v1.0 is a wonderful lora, it's fun to create something with it and i can't say what can be better in v2.0, not all pictures are top but the most are very good :) thank you for it



















