The goal of this LoRA is to depict full penile insertion together with the deformation of the surrounding flesh caused by close physical contact.In other words, it is designed to reproduce effects such as <code>ass ripple</code>, <code>ass press</code>, and <code>pussy press</code> resulting from <code>balls deep</code>. Initial release notes(v0.0.1)<hr />For v0.0.1, I trained the LoRA on 60 images in total:<ul><li>50 images showing full insertion to the base from various angles</li><li>10 cropped images focusing on the deformation of the surrounding flesh</li></ul>The overall result feels somewhat like a jack-of-all-trades, master of none.I compared the results with no LoRA, this LoRA, and LangdonAlger's balls deep LoRA. For full insertion to the base and completely concealing the penis inside the vagina, <a target="_blank" rel="ugc" href="https://civarchive.com/models/550870/balls-deep-deeper-penetration?modelVersionId=2885588">LangdonAlger's balls deep</a> still appears to perform better.For the deformation of the surrounding flesh, the Anima base model can sometimes produce interesting results. However, FT models often fail to reproduce the deformation. This is likely due to the insufficient number of cropped close-up images.<s>For the next version, I plan to significantly increase the proportion of cropped close-up images in the training dataset. Based on what I learned from developing my previous two LoRAs, increasing the proportion of close-up images should help the LoRA learn the deformation details more effectively.</s>As of v0.0.1, this LoRA still needs to be used alongside LoRAs specialized in localized deformation, such as an <a target="_blank" rel="ugc" href="https://civarchive.com/models/2625147/ass-ripple">ass ripple</a> LoRA.

Concept > Deep Penetration (includes ass ripple, ass press, pussy press)

Minor update to v0.1.0.The training images and captions remain completely unchanged. The LoRA was retrained with the following changes to the training configuration:<ul><li>v0.1.0: LoKr, dim32/alpha32, CAME, UNet LR 3e-05, constant scheduler, 1,000 steps, 80 training images, 512² resolution, 3 MB.</li><li>v0.1.1: LoRA, dim32/alpha32, CAME, UNet LR 3e-05, cosine scheduler, 960 steps, 80 training images, 1024² resolution, 66 MB.</li></ul>Results and Known IssuesThe higher training resolution may help recover details lost when the training images were downscaled, particularly flesh deformation caused by deep penetration.However, a side effect has been observed: in side-view and three-quarter-view compositions, the male character's POV crotch may appear even when a POV perspective is not specified.The latter half of the showcase includes comparisons between v0.1.0 and v0.1.1 across nine artists. Please refer to these examples when evaluating the differences.

Files with SHA256: 2f11ab7e2e2fd385b705e10d7ad30b1455bbd6ee9f6d2494b100bc6313898b39