LoRA Stacia - Goddess of Creation | Asuna | Underworld | SAO
Asuna in the guise of the goddess of creation Stacia. I tried to make the goddess qualitatively.
I will be glad to receive feedback for a better result!
This LoRA will be constantly improving for a better result.
Example of a basic outfit:
stacia, 1girl, solo, long hair, breasts, looking at viewer, smile, open mouth, bangs, detailed eyes, beautiful background, hands on the chest, brown hair, thighhighs, gloves, dress, cleavage, bare shoulders, brown eyes, very long hair, standing, full body, white gloves, white dress, armor, white thighhighs, garter straps, fantasy, white armorExample NSFW:
staciansfw, nsfw, 1girl, solo, long hair, breasts, looking at viewer, blush, bangs, large breasts, brown hair, thighhighs, ribbon, navel, brown eyes, sitting, very long hair, closed mouth, nipples, hair ribbon, yellow eyes, braid, nude, shiny, stomach, white thighhighs, shiny skin, parted bangs, detached collar, white ribbon, pink nails, white chokerAnother example:
stacia, 1girl, solo, long hair, looking at viewer, brown hair, thighhighs, dress, holding, brown eyes, very long hair, weapon, sword, armpits, holding weapon, white dress, armor, arms up, white thighhighs, floating hair, feet out of frame, holding sword, breastplate, fantasy, magic, energy, white armorFeel free to leave feedback.
For version 1.5
A completely retrained model.
A lot of work has been done to achieve the best result.
Must work with other LoRA.
Recommended parameters:
LoRA weight: 0.4-0.8
Sampling method: Euler a or DPM++ 2M Karras or SDE
Sampling steps: > 20
CFG Scale: 4-12
Upscaler: Latent or another
Denoising strength: 0.5
For version 1.1 - Unstable
Trained on the Anything-v4.5 model.
Recommended parameters:
LoRA weight: 0.6-0.8
Sampling method: Euler a or DPM++ 2M Karras or SDE
Sampling steps: 28
CFG Scale: 7
Upscaler: Latent
Denoising strength: 0.5-0.6
Hires steps: 0
For version 1.0 - OUTDATED
Trained on the Anything-v4.5 model.
Recommended parameters:
LoRA weight: 1.6-1.9
Sampling method: Euler a or DPM++ 2M Karras or SDE
Sampling steps: 28
CFG Scale: 7
Upscaler: Latent
Denoising strength: 0.55
Hires steps: 0
Description
For version 1.5
The model has been completely retrained. A lot of work has been done to achieve the best result.
Must work with other LoRA.
Recommended parameters:
LoRA weight: 0.4-0.8
Sampling method: Euler a or DPM++ 2M Karras or SDE
Sampling steps: > 20
CFG Scale: 4-12
Upscaler: Latent or another
Denoising strength: 0.5
FAQ
Comments (11)
Another Asuna fan I see. You my man, has good taste.
@f4ith Thanks! For the best girl, nothing is a pity.
So 1.5 doesnt work at all, i keep getting various errors, the latest one was this: NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.,
1.1 works just fine though, so im not quite sure what the issue is for the latest update?
I tested it in Google Colab and everything worked fine. Try updating the WebUI
What model do you use?
I read a little about this problem. Apparently, the problem is in Civitai. https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/6923
Try doing this: Settings > Stable Diffusion > Enable option "Upcast cross attention layer to float32".
@L115A4 I was using anything 4.5 my WebUi is not updated because an update a while back ended up breaking all my extensions so i reverted back to an old update.
@L115A4 Ill try this thanks!
Very nice Lora
However, I have noticed this lora like to have the arm up.
is there any way to negate this?
Thanks for the feedback!
I apologize for the long answer.
I didn't notice such a problem.
Try to explicitly set the desired arms position. You can also try using negative promts.
A radical option would be to use Control Net.
Most of the problems during the creation of masterpieces arise from incorrect promts
This doesnt work at all with the RadiantLight model no matter the weight its distorted the hell.
I dont have any other loras added either and still refuses to work
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.