Intended to produce POV anal sex scenarios.I had a dataset containing different kinds of positions, and at one point I ended up with different checkpoints that were better at certain things, so I will just have 2 model versions to download.It can produce the anus, but not reliably. Try it and see.Again, I trained this without captions, so your prompts do all the heavy lifting. If you find one that works better than my examples, please share it with us all.Prompt examplesFirst<pre><code>a pov view from above a cute young blonde lying on a bed, the woman seems petite, she is situated on her stomach with her back to the viewer, at the start of the scene the viewer is holding his penis, after a moment the viewer inserts his penis completely into the crack between her ass cheeks, he has some pubic hair visible above the penis, the viewer's penis appears slightly moist and darker where it penetrates her, the viewer's head is not visible in the scene only his torso area, the main focus is on the woman's ass</code></pre>Second<pre><code>a pov view from above a cute young blonde on all fours on her bed, the woman seems petite, she is partially undressed and on her knees with her back to the viewer, the viewer is sliding his penis completely in and out of the crack of her ass, he has some pubic hair visible above the penis, the main focus is on the woman's ass</code></pre>Third (with insert lora)<pre><code>a pov view from above a cute young blonde lying on a bed, the woman seems petite, she is situated on her stomach with her back to the viewer, at the start of the scene the viewer is holding his penis, after a moment the viewer inserts his penis completely into the crack between her ass cheeks, he has some pubic hair visible above the penis, the viewer's penis appears slightly moist and darker where it penetrates her, the viewer's head is not visible in the scene only his torso area, the main focus is on the woman's ass</code></pre>

POV anal (WAN14b)

This version is capable of producing POV anal inserting action, as well as prone scenarios.I trained using the other version as a base against a different dataset.

Files with SHA256: a2617ee65440d0eacae0e3ffccfff1b42a38ccbfc4105e119e7a06544eaf4665