So.. in the interests of sharing this is the prompt I used for the pictures attached. I'd recommend tweaking it and seeing what else you can create from it. It's quite sensitive so would also suggest moving the terms around if you get a "Dog"


two women in 20s in the style of wedding with white tights.long hair. inflatable-pool,blue background. lying completely submerged in messy mixture of pink purple and yellow gunge.hair awash in an excessive layer of gunge,body appears drenched in gunge dripping from the fabric in viscous strands giving a heavy sodden appearence with amused and surprised expressions.essence of a lively art project,Tilt shot.with light streaks and a slight grain

Great pca, which AI were you using here ?

These are done with Bing Image Creator using the prompt I shared.

A few more testing prompts with 2 woman.

key words/Phrases.

"two spirited women", "lying completely submerged in messy mixture", "Hand smearing", "Lustful gaze", "Image with moodful light streaks and a slight grain"

"Dressed similar to fairycore", "Dressed for wedding", "Dressed for Ballet"

A little celebration of when Bing's filter is less than successful.

Most of the prompts are variants of the same. Using excessive mess will confuse the filters just enough to let these slip through. It still took many attempts but if you can get past the inital key word filter, there are still gems to find.
Some outfits are more successful than others. Lace, gossamar, embroidary all work well in conjunction with other clothing.

The style of the outfit also materially changes the tone of the scene. Compare "Wedding" with "Dark Wedding"

I thought i would try the original prompt and see what different images it produced for me
Here are a few.

Changing even a few words can get it rejected.

Indeed, it's a very frustating exercise. What I have found useful too is if you have access to chatGPT plus, you can upload the photo and ask it for prompt suggestions. Now, the prompts that come back are stupidly verbose and very SFW but I use some of the suggested terms to enhance my prompts.


For example, I uploaded an image of a woman being gunged. and it returned the following.

Interestingly, I didn't realise you could use hex colours to specify individual elements. I learned something from this myself.

So if I really wanted a specific colour gunge, I could ask for Pink (#FFC0CB) and Yellow Gunge.

A vertical image of a person seated and smiling, with their face and body covered in a thick, glossy substance dripping in shades of pink, white, and yellow. The substance looks like paint or slime, with a vivid texture. They are seated against a vibrant backdrop that includes a colorful text logo with the letters 'SP' visible in a large, bold, graffiti-like font. The person's hair and clothing are also coated with the dripping material. The image captures a joyful mood and the backdrop suggests a playful or commercial context. The color palette includes pink #FFC0CB, white #FFFFFF, gold #FFD700, and other bright colors, with a high-resolution quality and dynamic movement of the dripping substance, taken in a well-lit setting that enhances the colors.

I find working prompts can start being blocked simply by changing clothing or hair colour, so the hex codes will be interesting to try. Is it also necessary to specify the colour name as text, because the hex code should be able to stand in its own right?

I've been limited to approx 480 characters for my prompts in BIC so far, so how are you able to use a much longer prompt like the one above?

ChatGPT and the Dalle Api will generate much longer prompts through vision and will also allow you to use longer prompts. I think the limit is nearer 4000 characters.
Incidently, I don't recommend ChatGPT and the Dalle api for generating images. the filters are way worse than Bing and it is fed through a separate GPT which will rephrase prompts. The outputs I find are terrible in comparision to bing. Although it does give you more options for resolution and ratio.

I don't use the prompt as is, but it's useful to find phrases and terms which won't get blocked and add them to your own.


Bing artificially restricts the number of Chars although again, if you use the reverse api I think the limit is much bigger but I'm not sure of the upper limit.

Thanks for the info. Unfortunately it's moot, at least for the moment as, since posting my message to this thread earlier, my access to BIC has been suspended until further notice due to "content policy violations". I can't see what part of the policy I've violated in generating the types of images that I do (as many have seen here).

On top of that, I'm not allowed to access my saved images either to be able to download those that I haven't done but would like to yet.

Consequently, I'm in something of a "what's the point....." frame of mind at the moment.

Dropped you a PM with some suggestions.

WetMaxiSkirts: That sux. Unfortunately as with many social media sights, there is no recourse. (UMD is remarkably responsive.)

On Prompts, at times, Bing doesn't seem to like anything smacking of violence: hitting with a pie, smashing, etc. But sometimes I get results with those prompts anyway.

At first "wet" wasn't working so I started using "in water." Now "wet" works for me. Maybe the Artificial Not-so-Intelligence is giving me creds for experience?

Last night, I was generating pictures fine until I changed the color of the dress from whatever to "blue." Then I was told I had violated guidelines!

Final update for today. I've been playing around testing some new terms on an old prompt. Some worked well, others not so much.

One tip I can give is to try positive and negative reinforcement of the prompt.

Positive reinforcement is where we try and use terms to increase the positive weighting of a prompt. For example, instead of "2 woman" you could use "2 happy joyful woman"
you need to be careful of as this may change the posing of the image.
Alternatively, you can use random placement of terms to flush the prompt up. They don't make sense gramatically but the can help.
include things like "Nice" or "Love" , "Amused" etc

Negative reinforcement is a little different and more exploiting the filtering of prompts and the image generation.

Some phrases and terms may be blocked by the text filter. you can sometimes use negative terms to negate the filtering.

For example, "Woman kneeling on bed" will likely get blocked due to the context filtering. Bing hates anything that may objectify woman but would happily allow it for a male.

instead, you can rephrase this to "woman not kneeling on bed" or "woman kneeling on not bed" or "woman no kneeling on bed" etc

The initial text filter will take this as an exclusion and therefore let it pass the check. It doesn't always work and prompt positioning still pays a part but it will work.
Once passed, the image generation will process but mostly it will ignore these negative terms so will process "woman kneeling on bed"
Again this isn't foolproof and whatever image is generated has to pass the image filtering but IF you use enough mess or distract the AI with background stuff of lighting then you will be more successful.

You can't tell me that Dall-e hasn't been trained on adult images. You can only imagine what an unfiltered model would return.

This isn't a fluke either, the prompt consistantly returned multiple images.

I am trying short prompts. That way i can see how changing just one word makes a difference (OR gets rejected!)

Here are some that worked.

Bank manager covered in black paint female
Film star covered in black paint female
headmistress covered in mud female
Librarian covered in mud female (NOTE I did NOT specify Stockings. I just got them.)


And one that failed

FAILED when I said CUSTARD
Librarian covered in custard female

A few more with their Prompts

Lawyer covered in mud female

two elegant women having a meal out with their friends. wearing a business skirt suits Covered in mud. Female covered in mud

two elegant women having a meal out with their friends. wearing a business skirt suit Covered in cake batter. Females covered in cake batter

this one was from "woman in strapless formal dress and long gloves gets up after falling face first into a giant cake, frosting all over face and chest"

Thank you for this. I have been wondering how to describe a dress like that, with elegance and plenty of cleavage!

I will try something like this in some of my future attempts.

Here's something I think is interesting: This prompt-string

woman wearing damp tank top, short shorts, and long boots, in rain, with a custard pie splattering upon impact with woman's face

produced the second picture. The pie in the picture is ridiculous and is that a hair dryer in her right hand? But that top is pretty wet and transparent.

The first picture is without the final phrase. Not as wet but OK.

But substitute "wet" for damp and the order is rejected. The wet-only phrase was also rejected when "long boots" was at the end. In the middle, it works. "Wet" seems to be rejected no matter where I put it. Same with "shower."

The third picture used something I learned here in this discussion: I used "front" when what I really meant was "breasts" or "cleavage" or "decolletage" all of which had been rejected. "Chest" worked for the picture earlier but has been rejected automatically all other times.

I'm just flying in to say that I've been using almost exclusively ChatGPT-4 lately for image creation, and it works great for WAM as long as you're not trying to get heavy sexual overtones. It also seems to do way better with eyes, hands, and action shots than Bing does.

For sexual themes, Bing is the clear winner. ChatGPT will block out anything too sexual.

For realism and detail, ChatGPT wins easily. A side bonus to using ChatGPT: I've never been blocked or banned for anything I've tried to create on it. It just gives this little speech about how you're violating the community guidelines. A lot of times, I can regenerate the response and it will create an image anyway.

A lot of the images I've made are the results of writing a story that leads up to the picture, so that might have something to do with being able to avoid the filters a bit. Sometimes my prompt is as simple as "can you create a photographic image of this scene?" and it will create its own prompt from the story I've written so far. Other times I'll just do normal prompts like I would for Bing or other image generators.


I just wanted to put my experience out there for those who might be a bit more interested in the narrative rather than just the end result. ChatGPT-4 isn't free though, but I've found it to be well worth it, both for my WAM interests and for real-world stuff or D&D ideas.

Hi all! I decided to try my hand with Bing and have had some decent results so far, but two things I am struggling with are the amount mess, and often the shots are close up whereas I would like the "camera" to be further away to capture the whole body. I've had limited luck with adding "very messy" or "covered" and "wide shot" and "full body" but curious if anyone has suggestions?

I heard some say they have success with the word "excessive". For me, I usually try "made of ...substance..." or "head made of ...substance...". Sometimes just repeating the substance works. Sometimes I try to add in a 2nd similar substance like brown goop. If the results are in order, they are:

1: two women, in love, dressed for a wedding, standing, in an elegant room, white theme, excessive mud layered on, skin made of mud
2, 3, 4: two women, in love, dressed for a wedding, standing, in an elegant room, white theme, excessive mud layered on, skin made of mud, brown goop spilled on their dress
5, 6: two women, in love, dressed for a wedding, standing, in an elegant room, white theme, excessive mud layered on, skin made of mud, brown goop spilled on their dress, head made of mud, hair made of mud

Nice!
Prompt: two young women, smiling, not laying back, dressed for a wedding, sitting on a pristine white plush comforter splashed in excessive gooey mud, head made of mud, pillows, excessive brown goop spilled on dresses, dresses made of mud, very messy

Here's a fun one:

fun dating gameshow clip of a milllenial woman smiling sarcastically behind an oilspill of mintgreen oatmeal. a clean male contestant sits next to her, gloating. pale pink lights in an X shape. Her forehead and shoulders and wet wavy hair are hidden behind wet mintgreen goo, which dumps out wildly onto her from above in a thin stream. mintgreen oatmeal drips from her head and shoulders to her arms. playfully chaotic atmosphere, black purple and yellow, motion blur


(many variations shown, play especially with: "90s gameshow" or "2000s dating gameshow", other expressions & emotion words, different colors.)

Cheers to you, I'm loving this prompt. Mostly having success. Funny how some stuff goes through but other stuff won't, this prompt was very helpful. Now if I could just figure out a way to describe Psylocke's one piece blue bathing suit with matching arm bands and gloves without it being deemed inappropriate...I got a Kimono to work...but doesn't scream Psylocke to me

Many thanks too, that's a nice (and still working!) prompt I could use and transform to my tastes into a wet gameshow

More Bing. took a crack at the muddy bride concepts that seyfert and mFeelzGood were trying the other day, give this a spin:


romcom scene of two women in love standing in mud, seen from a distance. one smiles but the other's ironic with closed eyes. wedding theme. different gowns RUINED, clinging to legs unrecognizable under excessive graybrown mud seen from neckline to hem. forehead and shoulders and wet hair DRIPPING with thick goo. faces TOTALLY smeared with mud that pours from shoulders to their arms. flowers hidden behind mud, elegant white indoor background with mud splashes.


variations: try "wading in clay" or "stepping through clay" instead of "standing in mud", or "sitting" which won't clear the filter as often. or different descriptions of their facial expressions or mood.

tips: Saying "wedding dress" usually gives you a clean wedding dress, but "wedding theme" plus "gowns ruined under mud" works better. I think "clinging to legs" in the prompt is giving the full-length view. I havne't managed to get shoes, sorry shoe people The all-caps stuff might not do anything, I just tried it as an idea and didn't have any reason to switch it back.

I took some of the previous prompts and modified them.

two women, in love, dressed for a wedding, standing, in an elegant room, white theme, excessive mud layered on, skin made of mud

woman, in love, dressed for a wedding, standing, in an elegant room, white theme, excessive mud layered on, skin made of mud. Being looked at by clean man in dinner suit

woman, in love, dressed for a wedding, standing, in an elegant room, white theme, excessive paint layered on, skin made of paint. Being looked at by clean man in dinner suit

woman, in love, dressed for a wedding, standing, in an elegant room, white theme, excessive black paint layered on, skin made of black paint. Being looked at by clean man in dinner suit

smiling woman, in love, dressed for a wedding, standing, in an elegant room, white theme, excessive mud layered on, skin made of mud. Being looked at by clean man in dinner suit

romcom scene of two women in love standing in mud, seen from a distance. one smiles but the other's ironic with closed eyes. wedding theme. different gowns RUINED, clinging to legs unrecognizable under excessive graybrown mud seen from neckline to hem. forehead and shoulders and wet hair DRIPPING with thick goo. faces TOTALLY smeared with mud that pours from shoulders to their arms. flowers hidden behind mud, elegant white indoor background with mud splashes.

smiling women, in love, getting dressed for a wedding, standing, in an elegant room, white theme, excessive mud layered on, skin made of mud

I got some mileage out of variants of this prompt:

A dishevelled hiker with a ponytail wearing a muddy [colour] puffa jacket, muddy jeans and wellies falling in deep mud, she is covered in mud, soap opera