After many weeks of working on it, Embrace the Suck V1.1 is here. I've added new capabilities beyond v0.81. New trigger words added: milf, cumonass, cumonbody, creampie, anal creampie, vaginal creampie, latinfemale, sucking man's cock, licking man's cock, and stroking man's cock. If your hands and fingernails don't look right, you can try "perfect hands" and/or "perfect fingernails". This may, or may not improve things. I've include a .zip with a txt file that has all of the trigger words I used.I have other models which are available exclusively on SubscribeStar and Patreon. So if you like what I am doing, and want to support me, you can do so on SubscribeStar at <a target="_blank" rel="ugc" href="https://subscribestar.adult/warped-animation">https://subscribestar.adult/warped-animation</a> or on Patreon at <a target="_blank" rel="ugc" href="https://www.patreon.com/WarpedAnimation">https://www.patreon.com/WarpedAnimation</a>. I prefer SubscribeStar, because it is unrestricted and uncensored, whereas Patreon is not. However, I realize that some folks around the world find Patreon payment methods more suitable to their needs.<hr /><hr />V0.81<hr />Well...it took me quite a bit longer to get to this point than I thought it would. After a great deal of time, experimenting, training, retraining, and even writing some custom code for my local kohya_ss software, I finally can release something.The blowjobs, handjobs, cumshots, cum, facials, etc..., are pretty much Beta quality (I think). I have also added in some additional sex capability, but I would call that portion of the model more along the lines of Alpha quality. With the right prompting , you may be able to get some female masturbation, and perhaps some sex positions out of it also.**** The focus has been on blowjobs, handjobs, posing, and various cum scenarios. Consider the other stuff a bonus for now.<hr />V0.71<hr />Update: I noticed in my testing of this version that profiles aren't coming out great, and that some other faces at a distance are weak. As a result, I've added in some more profile images and headshots to hopefully take care of that. Also, male genitalia at a distance doesn't appear to be coming out the way I'd like yet, but it used to be fine, so I'm guessing that the additional training will fix that, and that it is probably caused by all the additional data. This version was only trained for 8 or 9 additional epochs past that of V0.65.I've focused on hetro porn. However, the model IS capable of gay porn. Others have provided some examples of that in the comments for this model.Based on a somewhat creepy review that I received on this Version, I'd like to clarify something. This model is being trained on Adults, with absolutely ZERO focus on children. The fact that I have to state that, boggles the mind.Also, feel free to offer constructive feedback.<hr /><hr />Note: All sample images were generated with either ComfyUI or A111. In the case of A1111, I have indicated so just before the positive prompt used.I found that for the cumshot trigger word, sometimes a Clip Stop of 2 seems to work better.<hr />Barring something unforseen, this should be the last Alpha release. I will be locking down the training dataset to what it is now, and just training for several days. You should find this version much improved from v0.65. I removed some poor quality images from the training data, replaced them with some additional HD images, and added a full set of B&amp;W versions of the entire color image set. This doubled the side of the training dataset, so it now takes right around 5 hours per Epoch, as I'm sitting at just under 11,000 total images in the training data.V0.65<hr />Note: All sample ComfyUI images were often generated using Dpm ++ 3M SDE Eponential, but there was no choice for that in the list, so I just picked whatever was closest.<hr />I've added in additional porn capabilities. However, as the number of images has grown, the training time has grown to a point where I've decided to make this model about blowjobs, handjobs, facials, cum on breasts, erotic posing, etc..., and after I'm finished training this model, I will create another one to accomodate other porn. In fact, it might end up being multiple models, depending on my patience levels. That said, this version has added handjobs, blowjobs, cum on hands, black females, white females, redbone females, and some additional images on the pre-existing categories. I have trained this version a great deal, and it will require even more training before I will consider it a "Beta" or a "Final" version. Also, to correct some shortcomings of the trained model, I have once again merged in a very small amount of another model. Sample images were generated using ComfyUI or A1111. In the case of A1111 images, I've indicated such just before the positive prompt.V0.5<hr />Note: All sample ComfyUI images were generated using Dpm ++ 3M SDE Eponential, but there was no choice for that in the list, so I just picked whatever was closest.<hr />Additional training data and many additional epochs of training, and I'm releasing a second Alpha. This version contains additional capabilities, as well as better male genitalia. In order to solve an issue with hands being often times messed up, after training, I added a very slight merge of another model, and it seems to have mostly fixed the malformed hands issue. Sample images were generated using ComfyUI or A1111. In the case of A1111 images, I've indicated such just before the positive prompt.V0.3This is an alpha version of my "porn" model. Some male genitalia comes out half way decent. The "bodily fluids" look pretty good so far. It is not yet all encompassing (thus the Aplha versioning). This was trained using my RsmGalaxyRealCartoonXL_V1 model as a base, so the non-porn images should still work well. It takes many images and long training cycles to accomplish decent results, so I will be releasing updates to this model as I go, and with new capability.

Rsm Embrace The Suck XL

Barring something unforseen, this shouild be the last Alpha release. I will be locking down the training dataset to what it is now, and just training for several days. You should find this version much improved from v0.65. I removed some poor quality images from the training data, replaced them with some additional HD images, and added a full set of B&amp;W versions of the entire color image set. This doubled the side of the training dataset, so it now takes right around 5 hours per Epoch, as I'm sitting at just under 11,000 total images in the training data.

Files with SHA256: d44919b9b68225f1b5a009011629dca49b9ecfa5657efee5f4fc4d58bbbeb461