AI generated fetish images

[Like Ra]

This is truly the stuff of nightmares!

I thought you like powerful muscular women 😋😉

[Bound Whore]

Can you add a well pronounced erection here? 😋

Not convincingly, no 😉

[Bound Whore]

Can you add a well pronounced erection here? 😋

Well, saturday night and I had a few, so I gave it a shot.
Another practical life skill acquired ✔

[Like Ra]

COOL! 🤩 What was the prompt?

[Bound Whore]

COOL! 🤩 What was the prompt?

Oh, this was painstakingly photoshopped.

I tried and fed this as source image into the AI to see if it can get something from it by doing img2img.

Sorry, "Hermaphrodite, trans woman, visibly outlined penis bulge" are just not concepts that the AI has internalized.
YET!

Edit: Did I speak to soon? It really gave it all it got here.

[Like Ra]

Is the AI becoming sentient? 🤔

Are there any instructions how to train the pretrained model? Without breaking it, of course. The learning rate should be very low in this case, but the impact of the new data will also be very low.

[Bound Whore]

Are there any instructions how to train the pretrained model? Without breaking it, of course. The learning rate should be very low in this case, but the impact of the new data will also be very low.

With the version I have installed, I could theoretically feed it little batches of images (3-5 is the recommended number of images) and provide textual descriptions of what I want the AI to recognize in them.

I would love to try that, but this is quite memory intensive and I get an out of memory error with my 8GB GPU.

People who are more incentivized than me are renting out cloud computing to do this on a larger scale, but at around $150 an hour that's a bit much for trial and error runs.

There are aeady model data packages that can be linked in parallel to the official main model data package (which, if I remember correctly, cost about 600k to train), for instance there is a "Waifu"-package that is very anime-oriented.

It's only a matter of time.

[Like Ra]

I would love to try that, but this is quite memory intensive and I get an out of memory error with my 8GB GPU.

Aha... My goal is to get a 12GB card, I wonder if it's enough.

There are aeady model data packages that can be linked in parallel to the official main model data package (which, if I remember correctly, cost about 600k to train), for instance there is a "Waifu"-package that is very anime-oriented.

Oh! This is interesting! Any links? I guess more memory is needed here as well.

With the version I have installed

You will need one of the following:
- An NVIDIA-based graphics card with 4 GB or more VRAM memory.
- An Apple computer with an M1 chip.

4GB should work, but @Anne said, that 4GB is not enough.

[Bound Whore]

4GB should work, but @Anne said, that 4GB is not enough.

They have tweaked it a bit and now 4 GB is sufficient to generate images from prompts (512x512)

Training requires way more it appears.

I have not tried embedding additional model packages, but googling "Stable Diffusion waifu" should point to the more popular ones.

[Anne]

I spent all day installing Stable Diffusion on a pc with an AMD card (8GB), it was working but during generation my vram (8GB) and regular ram (16GB) filled up and the process crashed. I'll try again in a few months/years when I get my hands on an NVIDIA card.