JAV waka_ninomiya

Imitation model. Modelled after Japanese AV star Waka Ninomiya (二宮和香).

Today we have an atheletic one!

This actress is known for her gymnastic poses, very refined posture and body flex. Try the LoRA and you will understand what I am saying. If you are into atheletic, fit and a little muscular JAV girls, this is one you cannot miss.

PS: This is the first model that I tried training for 15 epochs, instead of 10. I also had another version that I trained by setting max reso to 768, 768. I might release that version as well if its stable for use.

Recommended LoRA strength: 1.0 (0.8-1.2 are plausible range dependent on checkpoints; higher values lead to better facial features)

Main Prompt:

wakaninomiya_jav, breasts, long hair, brown eyes,

variations: hair buns, black hair, tied hair

postures: arm up, arms up, adjusting hair, arms behind head

facial variations: natural makeup, grin, teeth, parted lips

Checkpoints that work well for this LoRA:

1) Realisian, or any of the Magicmix variants. I personally find Realisian's render being closest.

2) ClearRealismMix, or asian-based/eurasian checkpoints.

3) PerfectWorld, or other photoRealistic Hentai models (just not European models; with the exception of PerfectDeliberate). A little hit and miss with the facial features, but overall still OK.

4) RealisticMix. Stable performance, not up-there in terms of facial resemblance, but still workable with re-seeds.

Note for achieving good resemblance:

Use Adetailer for baseline generation (at 512, 768 etc) only, try to control LoRA strength at 0.9-1.0, adjust your denoise accordingly. For this LoRA, recommended to use Adetailer at ~0.35 to 0.45 denoise strength for close-up, and 0.45 to 0.55 for full body.

For this LoRA, you dont need to set the prompts for Adetailer separately. I prefer to use just the main prompt, and turn ADetailer on, since you dont want to overload the makeup, this actress generally do very light blush and touch-on only.

Recommended approach: is to not use HiRes, but upscale it with img2img instead.

Use a denoise strength of 0.25-0.35, and turn off Adetailer.

Issues:

If you know this actress well enough, she has the infamous(or famous) japanese crooked teeth (according to google). This is not easy to reproduce, since SD models usually correct for these artifacts and align the teeth. I tried looking up for a teeth-based LoRA that can alter this, but to no avail. Perhaps there will come a day that we have a teeth LoRA/Adetailer to fix that!

Post-production edits/Reproducibility:

Some preview images underwent inpainting manually on the clothing part because of the nature of some of the training data. Prompt responsibly.

Happy prompting. And do share creations~