So, of late I've been a bit busy with other things in my life, which is why I've not posted anything over in the Platinum section.

I have done a few experiments with Stable Diffusion AI, which I have installed on my PC (I don't use online generators). I've been using a LoRA file (low-rank adaptation) For those not familiar with the concept, it's a small file containing training parameters which modifies the output of the image - such as adding cum to
a face, and there are several like that available. This is an example of a recent image I made - I really like how this came out:

red-cumslut.jpg

Anyway, one of the LoRAs I was using is called 'Spunkmouth', appropriately enough. It focuses on cum bubbles, but it has a fairly limited number of images that were used to train it (about 24 I think). So the results aren't as good as I'd like, so I decided to try to make my own LoRA, and see if I can do better (using a training utility called Kohya-SS).

So for the past few days I've been trying to find examples of cum bubbles, which is pretty time consuming. So far I have almost 100 images, but I could use a lot more.
These are some of what I have:

bubbles.jpg

As I'll be sharing the result with the FF community, I wanted to ask here if anyone has any other suggestions for images to add to the training dataset - the bigger it is, and the more diverse, the better, although image quality definitely matters. Resolutions above 1280px are preferable. Given that I'd have to edit every one of them, it might take me a few weeks to do, although the training process itself only takes hours (my video card is an older 1080 Ti). I've tried searching for cum-bubble images on here, but it's difficult to do.

I'm OK with anyone posting good examples here in this thread. Also I'm up to discussing AI facial art with other AI artists here as well.

-- Fractured