Images enhanced with AI

[Heykevin]

View attachment 1315214

These are great, I’m in the process of writing a story that features girls like these, “Busts” that are amputated, blinded, teeth removed, tongues snipped to make them longer, and made deaf and mute. They exist only to suck cock, and are fed intravenously and kept on a shelve when not in use.

 

[fallenmystic]

I concentrate on Tensor.art because it is easy to use and has bountiful resources and no NSFW restrictions. You can achieve some beautiful results with Tensor.art, and I do love beautiful things.

I freely admit that it isn’t so good for torture. There aren’t a lot of sites that are. Pixai.art had some good loras, but then it went non-NSFW. You can still access them if you log in through Discord, but then they only work with the anime style.

If you wanted good torture art, you could do something on Pixai.art and then clean it up with Tensor.art Img2Img. (The one on the left is Pixai, and the one on the right has been cleaned up in Tensor.) Of course, you would be paying for both sites, but if you have any experience with Tensor, doing things in Pixai wouldn’t be too hard.

I am only talking about basic procedures here, because that is where everyone starts, and if you have any tips or advice on advanced techniques, feel free to share it. I am still quite far from being an expert.

That's why I recommended Pony Diffusion. I usually try not to talk too much about AI workflows outside the main AI discussion thread because I believe there's no right or wrong way in artistic creation.

Also, I know many, if not most, people don't get past the stage of writing random things in the prompt and hope for the best. It is not the best way to create good kinky images, but I also understand not everyone is willing to spend time learning advanced AI tools or searching for sites that provide them.

But I felt Pony Diffusion is something you should try if you rely heavily on prompting but still strive to control the output. I briefly tried generating some torture scenes in Tensor.art now, and these were what I could generate by prompting alone without using any Lora:



All you have to do is select a Pony variant model and write the prompt using tags you find on the Danbooru site. For reference, the prompt I used to generate the above images is as below:

Positive:

source_real, score_9, score_8_up, score_7_up, BREAK, 1girl, torture, naked, collar, chained, arms_behind_back, whip_marks, highly detailed

Negative:

score_1, score_2, score_3, from_behind

And that's about it. No controlnet, no Lora, and no additional img2img runs.

I only briefly checked the site, but it looks like Tensor.art isn't a bad option for those who want a free AI generator. They even offer things like control nets and some inpainting features, which are absolutely needed to create anything more serious with AI.

But it can be enjoyable enough to generate simple kinky images like that randomly. For that purpose, I think Pony Diffusion could be the best option to achieve better images without having to learn more advanced tools or spending much time and credits to get something decent out of a sea of bad or irrelevant random images.

 

[Faustling]

Nothing needs fixing with this one (1). It’s an excellent piece of work by troubleinterrortown in Deviantart (where you may find more of his work). I’m just using it as an example to illustrate how the models and loras work, and the variety of styles available.

2. Model: permissivebeauty_v30; Lora: Sherry Birkin . Resident Evil

3. Model: extra_realistic_plus_xl.fp16; Lora: Detail Master LoRA Pro Max(UNCENSORED)

4. Model: realisticComicBook_v10; Lora: Tomb Raider I Classic Lara

5. P@ny – Pony Mix; Lora: Skyrim female Breton XL

Half the fun is experimenting. Anyone care to express a preference?

 

[petelobo]

Something other than boredom on their faces and some marking on their bodies. Oops, I forgot, this is AI.

 

[HBDancer]

 

[Faustling]

Something other than boredom on their faces and some marking on their bodies. Oops, I forgot, this is AI.

I asked about the style, not the content. You can do everything in AI.

 

[Faustling]

A small sample of the kinds of images available from Tensor Img2Img. It's a powerful tool, so use it wisely.
This function requires very little instruction or experience to use. The main difficulty is identifying the best models and loras for your concept, but the learning curve isn't all that steep.

 

[HBDancer]

 

[John Hansson3]

I do really love these AI illustrations, thanks! They are often more real than ordinary drawings imo. Does one need any particular hardware to make them? And which app is best for torture pics?

If someone could make some AI electro torture pics and post here I’d be delighted. Thanks!

 

[fallenmystic]

I do really love these AI illustrations, thanks! They are often more real than ordinary drawings imo. Does one need any particular hardware to make them? And which app is best for torture pics?

If someone could make some AI electro torture pics and post here I’d be delighted. Thanks!

I consider 10GB VRAM to be the minimal hardware requirement. While running it with a less powerful video card is possible, it'll probably be too painful to use for any serious work.

That said, you can use one of the many online options to run Stable Diffusion, which doesn't require a high-end graphics card. What particular service or application to use will largely depend upon what you want to do with it and how much time and money you're willing to spend.

Some of the options are free and very easy to learn but can only generate simple images. On the other hand, there are more complex and powerful tools, which may require more time to learn and often a paid service (to clarify, most of such tools themselves are free, but you may need a paid cloud/host service to run it).

 

[John Hansson3]

I consider 10GB VRAM to be the minimal hardware requirement. While running it with a less powerful video card is possible, it'll probably be too painful to use for any serious work.

That said, you can use one of the many online options to run Stable Diffusion, which doesn't require a high-end graphics card. What particular service or application to use will largely depend upon what you want to do with it and how much time and money you're willing to spend.

Some of the options are free and very easy to learn but can only generate simple images. On the other hand, there are more complex and powerful tools, which may require more time to learn and often a paid service (to clarify, most of such tools themselves are free, but you may need a paid cloud/host service to run it).

Thanks! I’ll give it a try anyway!!

 

[HBDancer]

 

[Faustling]

If someone could make some AI electro torture pics and post here I’d be delighted. Thanks!

With AI, there are several ways of getting what you want:

• 1. An original composition, made in Tensor.art. You can make almost anything there, but it may take awhile to find the model that gives the best results. In this case, I used the GreyColorCuteMix – 1.0.
• 2. A piece of Hentai I found on Pixiv. The artist’s name is アヒル隊長 and you can find more of his stuff there.
• 3. My modification, made with the Tensor Img2Img function. In this case, I used Extra Realistic XL -v1.
• 4. A piece of hentai from Rule 34. Artist unknown.
• 5. A modification made with 123rf.com.

 

[HBDancer]

 

[Faustling]

If you reduce the number of images in Tensor.art from 4 to 2, you can use ADetailer, which as the name suggests, adds fine details. This is particularly useful if you wish to blow it up. The original images are on the left, the detailed ones to the right. Reduced in size for display here you may not notice the difference, but it is there. (Both sets use the EpiCRealism - pure Evo model, and the other settings are unchanged.)