Imitation model. Modelled after 0ppai-Goddess Japanese AV star Asuna Kawaii (河合あすな).

This actress is young, energetic, cute, sexy, fit and simply entertaining, its everything in a package (B88、W59、H89). She is well known for topping the DMM chart back in 2019 for the AV-debut newcomers category, and was also since then coined the G-cup breast goddess/Shin-yu (actual title 平成最後の神乳) within the JAV industry.

~~~~~~~~~~~~~~~~~~~~~~~~~~
Recommended LoRA strength: 0.8 (0.6-1.0 are plausible range dependent on checkpoints; higher values >1.2 lead to sharp facial features, but might also elongate and distort body proportions)

Main Prompt:

asunakawaii_jav, breasts, brown hair

variations: short hair, tied hair, wet hair

facial variations: parted lips, smile, teeth

*Usually, you dont need "short hair" by default, thats absorbed in the activation token.

and yes, its kawaii (pun totally intended)

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Checkpoints that work well for this LoRA:

1) Realisian. Models the facial features well. tend to generate a cuter version of her.

2) RealisticMix. Comes right after Realisian. Does better with generating the right body built, but facial expression can be abit more stiff.

3) Perfect World/Ultimate Hentai Realistic. Does OK, with occasional bad seeds (~20%).

~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Note for achieving good resemblance:

This LoRA actually require very minimal tweaking, atleast for this version - since I retrained the model after throwing out some training images that is generating very long torso for her. Thats fixed now.

To break this down, there's two versions of Asuna:

1) the smiling, cheerful Asuna: you will need to use "parted lips" or "smile:0.3" in your main prompt.

2) the default, gravure-like Asuna: just use the basic prompt outlined above.

Hint: Be mindful on the "smile" tag in the prompt, as some checkpoint tend to over-exaggerate her facial expression, and it gets alittle maniacal. I find one quick getaround, is to just use parted lips, and retain "smile" for Adetailer.

Depending on your checkpoint, you might need to set the prompt for face Adetailer separately.

Adetailer

Use Adetailer for baseline generation (at 512, 768 etc) only, try to control LoRA strength at 0.8, adjust your denoise accordingly. For this LoRA, recommended to use Adetailer at ~0.4 denoise strength for close-up, and 0.45 to 0.50 for full body.

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

Use a denoise strength of 0.25 -0.3 (anything higher will overcorrect face during the process), and turn off Adetailer.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Issues:

No issues so far for verion 1b, which seems to fix the long-torso problem in v1 (not released).

in img2img upscaling, sometimes the shape of her eyes can be alittle off - I guess you can try running Adetailer again with very low denoise to try and fix that.

Post-production edits/Reproducibility:

Actually none. though png prompt might not correspond well, since I upscaled some of these via img2img using a different checkpoint, accidentally.. But they shouldn't be too far off.

Happy prompting. And do share [SFW] creations~