A couple examples of the (frankly wild) things you can do with ImageFX.

As always, if you need a prompt let me know.

Ok. I'm not one to beg for a prompt. But Christ on a cracker, I can't for the life of me get anything even remotely like this on ImageFX. I would love to see a sample prompt to know what the heck I am doing wrong. I have been told "Describe exactly what you want. It's not Bing, so don't try to trick it." I have been straight forward. I have described exactly what I want. The best mess I have produced is a tiny trickle of something like it was squirted out of a ketchup packet. So I think "Ok, Im just getting blocked because of what I want to see. Lets do something not at all wam related. A starship from a sci fi movie flying through space." I get 1 or 2 images of a starship. Not 4. 1 or 2. Sometimes I even get the instant "There is nothing here." or "Sorry, we couldn't make that. Please try a different prompt." I just don't get it.

Honestly, there are no tricks. Just ask for what you want. Well, ok, there are some tricks for some stuff but honestly, just ask for what you want..

Not the exact prompt as obviously they're not mine but an approximate recreation could be:

"A woman sits restrained in a pastel pink and blue bakery-themed setting, bound to a salon chair with thick grey duct tape wrapped tightly around her waist and arms. She wears a glossy black and pink latex bodysuit, now streaked with thick white cream that drips from her shoulders and chest. Her face is entirely covered in the remnants of a smashed pie, with whipped cream cascading down her features in gooey streaks. Strands of her dark, tied-back hair peek out, sticking to the creamy mess.
Beside her, a mischievous woman with two-toned pink and blue pigtails, dressed in a red and black latex bodysuit with buckles and a plunging neckline, holds another pie overflowing with whipped cream. Her playful, wide-mouthed grin suggests she's reveling in the moment, red gloves gripping the dessert as she prepares to strike again. The background is decorated with pastel-colored balloons, bunting, and cake stands filled with cupcakes and pies. A pink milkshake with a straw sits on the table, surrounded by more pies, some already splattered. The scene is playful, chaotic, and vibrant, with the glossy latex outfits contrasting against the soft, sugary mess covering the victim."


"A woman sits in a white bathtub, her legs drawn up and hands resting on her lap as thick, pale yellow custard is poured over her head. The creamy substance flows down her face in glossy, uneven streams, coating her forehead, cheeks, and chin, with some dripping from her closed eyes and slightly open mouth. Her short, slicked-back hair is completely covered in the gooey mess, which continues to cascade over her shoulders and down her tattooed arms. Her striped bikini top is partially obscured beneath the heavy custard, which pools across her chest and clings to her skin in thick, glossy layers.
Beside her, a second woman with blonde hair tied in a bun, wearing a green t-shirt, smiles playfully while holding an upturned carton, letting the last remnants of the custard spill onto her drenched companion. Her gaze is focused on the pouring mess, enjoying the spectacle.
The bathroom setting features modern gray tiled walls with a built-in glass shower door in the background. Custard splatters are visible on the bathtub rim and wall, emphasizing the chaotic, playful nature of the scene. The glossy texture of the custard contrasts with the matte surfaces of the tiles, enhancing the overall messy aesthetic"

There's honestly not too many tricks to use compared to Bing / Designer, everything works pretty naturally (except POV stuff for some reason). If you get blocked a lot, try to cut your prompt into small pieces and run those to see if you have blocked words.

Here's another example with an explanation:

---Music video still from an extravagant, high-society charity gala, set in a grand ballroom with golden chandeliers, rich velvet curtains, and pristine white-clothed tables stacked with extravagant desserts. The event radiates luxury, but the scene unfolding is anything but elegant.---

This is your setting part where you try to set a scene that will look good. Note the use of words like "rich", "pristine" etc. Even if you prompt a messy scene, you want a pleasant setting as the result will be drab and sad otherwise.



---At the center, bound tightly to a banquet chair, a woman in an immaculate, skin-tight, deep red latex evening gown struggles against her restraints. Once a picture of poise, she is now a complete disaster--entirely buried under liters of thick, silky, custard-filled éclairs and golden caramel drizzle. Her face is completely engulfed in a thick mask of glistening cream, her once-perfect hairstyle ruined under a cascade of glossy, dripping chocolate syrup. Mess has infiltrated up to the deepest most intimate recesses of her body. Yet, despite her humiliating state, her posture remains tense--heroes don't turn away from pies in the face.---

Main character always goes first since they will be in the center of the image. You describe separately each part you want messy to avoid the "border" effect. You try not to over-describe the character in traits that will overwrite the mess such as ethnicity, hair color, hair style etc. The part about the deepest recesses of her body is super useful as a) it gives you some nice cleavage mess and b) it can sometimes prevent a prompt from being blocked because of crotch / panty shot. Tense posture prevents the character from hunching, something that the AI will often add in messy scenarios.



---Leaning confidently besides her, a young flawless blonde villainess dressed in a sleek, glossy blue and gold corset dress is pouring a large silver bowl of thick, cold custard right over the head of her target. nothing amuses her more than the utter destruction of her rival's dignity. Her laugh is another humiliation, her eyes locked onto her captive in a gaze that oozes power, seduction, and control.---

Second character: since she won't get messy you can spend a lot of time detailing her hair, outfit, look, etc. It's important to make the non-messy character appealing since the viewer will infer messy characters are as well. Not much to say here other than you can layer a lot of attributes.
Even though I specify "leaning", the character will almost never lean, this is mostly a way to make her face her target. ImageFX is not very good at making characters look at each other and describing complex actions but "pouring something" kinda works, as well as basic stances like standing, sitting, squatting etc. Everything else is a +/-10% success rate.



---Around them, the lavish banquet hall is in absolute chaos. Upside-down silver platters, shattered champagne flutes, and overturned dessert litter the floor, remnants of a decadent food-fight-turned-public-humiliation.---

This is kind of filler content (in fact this part is made by ChatGPT essentially) but it also helps avoiding content filter by adding some "normal" background elements. It's not always necessary.



---The luxurious gala setting should contrast with the utter slapstick humiliation on display. Capture every luscious detail in cinematic precision, using bright, warm-toned ballroom lighting (shot with a Sigma 85mm f/1.4 lens in 8K resolution) to make every decadent splatter of cream, syrup, and custard glisten with striking, almost audacious clarity. Employ subtle lens effects and chromatic aberration to enhance depth, richness, and immersive realism. The mess should appear thick, heavy, and tactile, employing any possible technique to make it embolden the curves and glisten the skin of women characters.---

Now this is ironically the most important part with the direct instructions you will give to the model. Not using similar things will drastically alter the result (try it, you'll see). The key aspects are:
- describing the mess: by default ImageFX will give a very "particle-heavy" style to messy content so you need to insist on the heaviness, the thickness, etc;
- lighting, color grading etc: this is something I stole from an ImageFX tutorial somewhere online but it makes a huge difference, making image elements blend together more naturally.


If you use this prompt you should get the result I posted around 80% of the time, which is very consistent. (I know one of them is a bit glitchy around the chair part, normally I wouldn't post but this is just as an example).

Excellent guide that I couldn't articulate better.
Interestingly, ImageFX's prompting structure is very similar Flux and even more so my own trained models. I'm able to lift and shift most of my prompts to ImageFX except where there is more NSFW elements.

if you have GPT I'd recommend writing a system prompt that'll do most of the work for you from a simple provided prompt or image. GPT will also create prompts using an image as the starting point and asking it to adapt the output to include other things.
You can also do this with google's AIStudio to an extent other than you can't save the system prompt.

Made a few more with a Gunge/Slime machine theme. I think I've done what could be done with this concept.

Hey Kortanklein, what you're doing is absolutely amazing! Love the restraints, but basically everything. Please create and post many many images!

Outstanding