The modern appropriate version of this is Pirate or l33tspeak, it's a common technique for debugging localization issues without having to be able to read another language

Don't downvote just because you can't read [T]ags, folks

I think ethics will also have to grapple with the fact that the thousands of humans who created the training data, often without their knowledge or consent, aren't compensated and not only didn't consent to their likeness or work being included but also can't effectively request to be excluded. I see plenty of Emma Watson or other celebrity lookalikes in the same models I've been using, and I don't think you can call yourself ethical while using tools that contain such stuff. No more ethical than the ethics of a person choosing to take a nude photo of themselves and charge money in order to see it, certainly.

Thank you for taking a stand against transphobia and abuse! Your efforts are valued.

Here's some bonus ones since people liked them

Oh trust me this was the best of four lol

It's hard to get right sometimes, like I'll often use the doggy lora just to get realistic body parts but then it tries to generate buttholes on top... can't win lol

Thanks! I keep trying to get huge boobs without it but it's hard to defeat the law of averages

With Deliberate (the spice must flow!)

{
  "seed": 2537105551,
  "prompt": "Photorealistic, hdr, 8k, cinematic, woman, body piercings, blue glowing eyes, large eyes, large iris, sclera, tan skin, looking at camera, queen, regal, jewelry, textured skin, freckles, incredibly detailed",
  "negative_prompt": "(deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation",
  "cfg_scale": "7",
  "width": 512,
  "height": 768
}

Sure! LoRAs are key, and you bet I deleted 75% of what I got.

{
  "seed": 2521279143,
  "prompt": " realistic photo of many men and women having groupsex, orgy, sex, nude, beautiful, dramatic lighting, 8k, vaginas, penises, penis, vagina, cum, glans, (penetration, intercourse:1.5), labia, cinematic, Kodachrome ",
  "negative_prompt": "fat, paintings, bad face, bad teeth, bad nipples, sketches, (worst quality:2), (low quality:2), scary, disturbing, (normal quality:2), lowres, ((monochrome)), ((grayscale)), bad anatomy, DeepNegative, deformed, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, bad feet, cropped, poorly drawn hands, poorly drawn face, mutation, deformed, extra fingers, extra limbs, extra arms, extra legs, malformed limbs, fused fingers, too many fingers, long neck, cross-eyed, mutated hands, polar lowres, bad body, bad proportions, gross proportions, text, error, missing fingers, missing arms, missing legs, cuts, injuries",
  "cfg_scale": "8.5",
  "width": 512,
  "height": 768
}

as you can see the random seed has about as much to do with the result as the prompt sometimes. I think the seed gets stuck on an idea like a certain example of the scene and if you try to shoehorn in a result it'll just get mangled (a person looking to the left vs the right, for example, if the dataset didn't tag heavily for that)

I've also had exact prompts and seeds drastically change output on me over time (maybe anydream updates their models?) so in a few weeks these may not generate the same stuff

{
  "seed": 3232717257,
  "prompt": " realistic photo of three men and four women having groupsex, orgy, sex, nude, beautiful, dramatic lighting, 8k, vaginas, penises, penis, vagina, cum, glans, (penetration, intercourse:1.5), labia, ((perfect face, perfect teeth)), cinematic, Kodachrome ",
  "negative_prompt": "fat, paintings, sketches, (worst quality:2), (low quality:2), scary, disturbing, (normal quality:2), lowres, ((monochrome)), ((grayscale)), bad anatomy, DeepNegative, deformed, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, bad feet, cropped, poorly drawn hands, poorly drawn face, mutation, deformed, extra fingers, extra limbs, extra arms, extra legs, malformed limbs, fused fingers, too many fingers, long neck, cross-eyed, mutated hands, polar lowres, bad body, bad proportions, gross proportions, text, error, missing fingers, missing arms, missing legs, cuts, injuries",
  "cfg_scale": "8.5",
  "width": 512,
  "height": 768
}

These models have a lot of trouble with realistic vaginal sex, I think this one accidentally stumbled onto a clit piercing on top of its usual struggle to separate out vagina, clitoris, and labia. (Sometimes it decides it's anal for no good reason)

I only include on request because it's a chore and often the prompts only bear vague resemblance to what comes out and the prompts are often copy pasted from existing examples, or random. The docs.anydream.xyz site is real useful.

{
  "seed": 1265397458,
  "prompt": "Realistic beautiful woman, ((succubus)), ((nude)), large breasts, nude, legs spread wide, ((eyes closed)), ((open mouth)), perfect teeth, wet oily skin, facing the camera, (evil tattoos covering entire body:1.4), two arms, two legs, horns\n\n\\n<lora:POVDoggy:0.95> 1girls, penis, doggystyle, implied sex, cum",
  "negative_prompt": " flat-chest, skinny, fat, paintings, sketches, (worst quality:2), (low quality:2), (normal quality:2), lowres, ((monochrome)), ((grayscale)), bad anatomy, DeepNegative, low-res, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, (signature, watermark), usernchild, kid,",
  "cfg_scale": "6",
  "width": 512,
  "height": 768
}

Many weeks worth of free credits, a lot of trial and error, and a Character saved to keep the face the same.

Thanks!

This progression is a lot of trial and error and then either using a Character to keep the face or keeping the prompt very similar and/or using the same seed between images. It's real random chance with the seed if you get something nice or off the wall. I couldn't get the "green eyes" to not totally change the whole appearance so it ended up as a header image. I had to specify skin color and ethnicity in order to not get tons of variation for example, and the more you try to nail down AI often the slipperier it gets.

{
  "seed": 161256961,
  "prompt": "portrait of dark-skinned African super model elf girl, (long pointy elf ears), (((from_behind))), perfect body, full body, huge boobs, H cups ,huge bubble butt, (sexy), curvy, naughty, (thigh_gap), vagina, labia, booty, Dark, moody portrait, girl, striking features, deep shadows, low lighting, mystery, beauty, intricate details, dramatic composition, tension, darkness, skull, candle, motion, wispy hair, dark color palette, contrast, texture, realism, high-quality rendering, stunning art, high quality, film grain, Fujifilm XT3",
  "negative_prompt": "NG_DeepNegative_V1_75T,(((multiple heads))), (greyscale:1.2),disabled body, (worst quality:2), (low quality:2), (normal quality:2), lowres, normal quality, ((monochrome)), ((grayscale)), skin spots, acnes, skin blemishes, age spot, glans,",
  "cfg_scale": 9,
  "width": 640,
  "height": 800
}

This was unintentional lol the random prompt had "candle" towards the bottom

Absolutely, neural nets are mind-blowing even if they run into obvious limitations

If I could get this damn program to generate fangs or zombified skin I might have something besides horns and blood to go on lol

Never understood the appeal of using AI to make an image of something that could exist in real life, that's just paying the wrong people at that point ;)

Thanks! That part is random so thank AI and the human artists who created the originals

Look at http://docs.anydream.xyz you need to use a lora and trigger it with the activation keywords. I believe the words should be after lora:example:0.9 in order to take effect.