Cum isolated as a material. Cum on whatever you want. Cum super zoomed in and detailed. An object cumtagged tagged with cum. Cum as letters? Yes. Cum as a lifestyle choice.
Sure, there are many cum loras for Krea 2, but this one is different. It is specifically focused on making super highly detailed and realistic looking semen, jizz, spunk, what have you... on objects.
If you're looking for just the cum on a surface, this is the right one for you.
It's best used with LLM-made prompts, with gemma 4. The best results come from using thick, white, viscous cum on your object of choice.
It is trained for the word cum and should be flexible around that. If your prompt swaps cum for semen or any other word, it just won't work as nicely. Make sure it's cum.
Can be used at 0.6-1.0.
What this won't do: cum in a cup, cum far away (just looks like white glue).
I haven't tried it with faces, as that isn't the point. I bet it would do it.
This model understands what cum looks like, beyond just these examples. It can be made to make manmade horrors beyond comprehension.
Description
FAQ
Comments (3)
OMG perfect!! Now a precum would be the cherry on the cake. 😉
Probably the best semen texture of all semen LoRAs currently available for Krea 2. It is, however, only trained on things, not skin, and that shows. (Please don't take this as a complaint, as I know skin wasn't the focus of this LoRA!)
The problem is that it makes the skin under the semen turn daker (see gallery image by nekotaylor for example). I suspect that's a result of the training data. A lot of surfaces react this way to liquids, appearing darker when wet. Skin, with its naturally oil-coated outer layer does not, as that layer prevents the liquid from altering its light reflection.
On the plus side, since no faces were used in the training data, it doesn't change the character's facial features, makeup, etc, unlike almost every other semen LoRA.
Again, I know skin is not the focus of this LoRA, so I'm in no way complaining. Take this as my request for either further training on skin (without including faces in the training data!), or a separate version for skin.
I think you're spot on, this is definitely because of the training data. I've noticed a similar quirk when glass is involved. It likes to make whatever is under the cum turn dark and saturated, even if that's not how it should look in reality.
Thanks for the comment! If I do another version someday, I'll consider finding more data that shows the right way to depict skin.