Read Description

Thanks for the heads up! I must have accidently saved those in the wrong folder during testing.

Can you elaborate? The model is in safetensor format.

I'm not sure. With AUTOMATIC1111 webui, you can try adding the following arg to your 'webui-user.bat' file:

set COMMANDLINE_ARGS=--VAE-dir "C:\path\to\vae"

I've never tested this, but that's how you specify hypernetwork and embedding paths so maybe it will work for VAE?

Use attention/emphasis literals () []
For example, (wearing open panel lingerie:1.2)
Or find other words for that mean the same thing. So instead of saying "Open panel" or "ouvert", try "Crotchless".
You could also try combing words that relate like this: (wearing open panel lingerie, lace mesh, exposed ___)
My last suggestion would be to combine your main focus (the person) and what they're wearing inside one literal. For example; A sensual photo, of (a woman, wearing open panel lingerie, lace mesh, Crotchless, exposed ___)

Fill ___ with whatever body part you want exposed, 'Open Panel' and 'Ouvert' could mean a few different things, which might be part of the problem.

Extra tip: Search the web for the type of lingerie you're looking for, copy the products name and paste it into your prompt; (wearing After Midnight Signature Lace Open Panel Teddy Bodysuit)

I actually have a wildcard named lingerie that is made up various lingerie products scraped from the web. Sometimes the model will generate the exact product that was put in as an input.

@Xeltosh No problem!
You need to install this extension - https://github.com/AUTOMATIC1111/stable-diffusion-webui-wildcards
Alternatively, https://github.com/adieyal/sd-dynamic-prompts is a more feature packed version of the first extension.

Unfortunately, I can't provide my wildcards here. They're quite easy to make though.
Example of making a wildcard for lingerie :
1. Create a text file named lingerie.txt

2. Inside the text file add some different types of lingerie. for example;
fishnet bodysuit

Bodystocking Lingerie

G-String

Lace Sheer Mesh Two Piece

Punk Leather Body Chain

Sexy Sailor Collar Sleeveless Backless Bikini Briefs

Short Mini Skirt

ect..

Note: Each lingerie set should be placed it's own line. So "fishnet bodysuit" is on line 1, "Bodystocking Lingerie" on line 2, "G-string" on line 3, and so on.

3. Save the file.

4. Place the txt file inside the wildcards extension directory.

In AUTOMATIC1111 webui the default file path is:

sd\extensions\stable-diffusion-webui-wildcards\wildcards\<place your .txt file here>

5. Launch AUTOMATIC1111 webui

6. Go to the settings tab and look for "wildcards". Here you can choose whether or not you want to keep the same seed for batch image generations. (This will make more sense in a moment.)

7. Go back to Txt2Img and write your prompt, instead of writing "wearing fishnet bodysuit". Instead write, "wearing __lingerie__".

8. This will pull from the text file you created earlier and insert a lingerie set. So "wearing __lingerie__" becomes "wearing Short Mini Skirt", or "wearing G-String".

Note: Being a wildcard, the selection is made at random.

In what way are models not working for you?

@nft_enjoyer Go to your settings tab in webui, look for a setting labeled something like "ETA Noise Seed Delta" or it might just be called "ETA". Set it's value to 31337

This is assuming you're using AUTOMATIC1111's webui. I can't confirm if this setting exists on other tools.

I just tested a few images with the copied generation data from this page and the generations were exact copies. The only difference (as of now) that I can think of, is the ETA value being different between my setup and yours. Hope this helps!

@nft_enjoyer That's odd, is CODEFORMER set to 0.8, and enabled for Face Restoration? If not, try enabling it. Also, double check to make sure the image sizes are identical. Size matters. I think most of the sample images for this model are 512x768. The generation data should give the exact size. One last thing would be to double check that this model is loaded in Webui. This models hash is: 7C5A435D50
So the model loaded (top left of webui), would look like <model.safetensors (7C5A435D50)>, or
<elegance_244.safetensors (7C5A435D50)>

If all else fails, I would need more information about your setup to help further.