Fire drills at my office were never like this...

(And I think someone needs to have a word with those two cops in front of the car about their irregulationary uniforms.)

I really like these from an artistic point of view.

More similar to umd wam sessions

Party hardcore
And corporate party

End of the year corporate party

Something slightly different -- my attempts to get Bing to illustrate a scene from Carter Dickson's 1943 detective novel "She Died A Lady".

I love the idea of car stuck in deep mud and getting messed up trying to get out, these are great.

In the book the car's sinking in quicksand and she has to jump clear before it goes under, but that was as close as I could get.

Bing is weird. Today I was working on a wetlook set. I had all of my dry images done, all of my underwater images done, and was trying to work on my wet surface pictures, and I got the dog 9 out of 10 times. The one picture I would get was low quality and not wet at all. I thought it had a problem with the outfit, so I went to another account and tested my prompts from another set that I knew worked, changed the outfits, and I got the dog. I changed outfits, dog. I removed the background, dog. (I remoced the background because the prompt that I knew worked technically didn't have one, so I reverted to that.) So then I just tried the prompt that I knew worked as it was, dog. Over and over again. Changed the gray t-shirt and jeans to a Supergirl costume, dog 7 out of 10 times. Changed the pool of water to pool of slime, 3 to 4 images every single time I hit the button. I tried going back to the t-shirt and jeans with the slime, and got the dog 9 out of 10 tims again. The 1 image I got back was good (to me anyway). So I thought I would throw that one, plus Supergirl getting wet and messy in here for you all. Couple clean-up images as a bonus.

Oh my gosh these are glorious

Can I ask how you do wet pictures

Sure thing. As an example:

close-up photo of a millennial female. coated in water. in a pool of water. in-ground pool. neck deep in water. splashing. water is dripping off the head and clothes. grey t-shirt, jeans, glasses, beautiful long blond hair smoothed sleek and flat, wet hair, wet clothes, summer day

So, first off, there is a bit of science to this, and a bit more luck. I was getting some amazing images with pretty much this exact same prompt two days ago. Yesterday it wasn't working so much. But Bing is like that from time to time.

Anyway, once you establish your model and background setting, you need to add a source of ongoing wetting. If you put the model in a pool of water, they will only be wet below the surface of the water. You need rain or something else to actively wet the upper half of your model. "Splashing" works to that effect. Sometimes it will act like somebody dumped a bucket of water on their head, other times it will just give a little ripple in the water and still soak the upper half of the model. You can also use rain to the same effect. The "coated in water" just kind of makes the splashing effect work. I have never gotten a model completely soaked without that prompt, but it also doesn't work alone. Like you have to have a good battery AND gas in your tank for your car to run. The prompts "wet hair" and "wet clothes" are not always needed, but give better results more frequently if they are there. The "hair moothed sleek and flat" bit helps a lot too. Styled hair doesn't want to give the fresh out of the water look anywhere near as often. The "wet hair" prompt just makes hair slightly damp usually. The combined source of water, the dampness, and the hair being straight all combined give me the most soaking wet hair results. "Wet clothes" is another one that is kinda meh. I never notice it doing anything without the other prompts, but I believe it returns more good results than just the other prompts alone. Then you just figure out your posing and stuff. These are the tools in my belt that I use to get my results. That's the science. I would say it's still 50% or more luck just hoping Bing plays along with you at that time.

All of these images have used the prompt I shared in this post. I just changed the outfits, posing, and background.

Thank you

Added cars and general mayhem... it appears as more crazy stuff lets other stuff slip through. Also, adding cars gave wide angle shots. No other prompts did that...

I also played around with "nationalitites". Using e.g. Dutch created one set, but using "Russian" not only created other types but added crazy props (like the, it looks like a rifle or something)

Yeah, that seems to be the case. I'm doing my best to not have a lot of crazy and weird stuff going on in my images, and I'm paying for it dearly. I also had a crazy streak of luck by not including a background at all (in my head, I'm calling it "the test room" lol). I'm finding it very difficult to produce even remotely the same images in a tame setting like a pool in a fenced in back yard. At least im only getting the dog and not prompt strikes.

I read a thing the other day that said something like "The internet is 95% porn, so AI WANTS to make porn". I can see that. Look at my images above. The one where the shirt is falling down past the shoulders and damn near exposing the breasts. I didn't do that. Bing did. We are getting blocked on SOOO many images mostly because the AI is dirtier than we are. In a perfect world, we would be allowed to make our own decisions and turn a content filter off. Make us pay for it, restrict images so you have to download them or lose them forever when you hit create again. Whatever. But I can't stop thinking about how many fantastic images have been created and then thrown in the trash by that damn dog without us ever getting to see them.

I thought about it already before.
Most likely that you are doing advertisment for insurance company.

"an athlete in running vest and leggings crawling through a muddy puddle"

"a female cyclist wearing a red lycra outfit covered in mud, riding through a muddy forest"

Playing with some neon lights

This is perfect

That's what you can do in other AIs like Stable Diffusion, there's usually no prompt restrictions at all. On the downside Bing/DALL-E is far superior in many things like image generation and usability.

In a perfect world we could have a blend of all those systems, the simplicity and generation capabilities of DALL-E combined with the flexibility and uncensored possibilities of Stable Diffusion and the realism of Midjourney.

Sometimes I also get good results with prompts like "after being dunked", "after diving" etc. depending on the rest of the prompt and scenery. Sometimes it will also accept things like "soaking wet woman", but this may as well get blocked by the dog or result in a content warning or temporary lockout.

PLayed around a bit with midjourney v6 today ... it's still alpha, but the level of detail is already insane. I guess they will ramp up their filters in the coming days so have fun while you can ...

Just a heads up regarding image designer if you've been using that recently.
There has been a significant update over the last day or and the prompt rejections.

basically , designer got updated on the back end, it seems. if you submit more than a handful of prompts within an hour or so, designer will hit you with the "we're having a lot of visitors right now" block. previously it was a 60 second cool down, but now its far more aggressive, and locks you out for anywhere between 10 minutes and an hour. there is also a theory that the more of those you hit, the longer the cool down gets, with it resetting every 24 hours or something. the first few gens you do each day seem to go through easily but then designer slows you way down.

Similarly, if you receive "Images couldn't be generated......" or the "It's us not you" prompts you need to wait a significant amount of time before you try again or risk getting the "We're getting a lot of visitors" message

This is making Designer pretty unworkable at the minute to produce anything significant.

A few Wedding themed pics

few more