v1. Stable release @ epoch #8. trained using 99 High Quality images.










OK, many of you probably won't know this one. Mika Kayama's debut in AV is fairly short-lived, having started as a gravure model in ~2002 under the alisa of Arisa Oda (小田有紗), and only joined the JAV industry for 4 years before retiring in 2011. Some of her work and appearance do leave an impression among fans, especially those that follow the early days of JAV company s1 very closely. I personally find her petite physique and square-ish facial structure within a mature JAV genre quite unique, hence this model.
variations: black eyes, medium hair, tied hair, hair bun, earrings
facial variations: parted lips, lipstick, lip gloss, blush
*Use of facial prompts depend on checkpoints, use it if your checkpoint have a bias towards minimal makeup, most of the time you wont need any of these
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Checkpoints that work well for this LoRA:
1) Realisian/RealisticMix. Back to the almost-perfect duo. I dont even need to cherry-pick, most pictures are quite on-point.
2) PerfectWorld and UltimateHentai My only complaint is the more 2.5D style render. Nothing major.
3) Other Asian-based photorealistic models. Only tried those, doesnt sway too far. Technically even merged models like GrayColorSemiReal-mix and ClearRealism-mix works well too.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Note for achieving good resemblance:
This LoRA is easy when it comes to reproducing good resemblance. 2 Steps:
1. Generate the main prompt using a LoRA strength of 0.6~0.7, higher for full body modes.
2. Supplement that by enabling face Adetailer, and set the weight slightly higher to 0.8. e.g.
"face of mikakayama_jav, long hair, brown eyes, parted lips, lipstick, <lora:JAV_mika_kayama_v1_epoch_8:0.8>"
The makeup prompts are optional, but can help create a more "mature female" appearance, if you wish.
It is still better to use AD at baseline generation (512,768 etc) and then upscale it later.
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:
BadHands mainly. Nothing major.
Post-production edits/Reproducibility:
Minor inpainting on some Badhands, and clothing.
Happy prompting. And do share [SFW] creations~
v1. Stable release @ epoch #8. trained using 99 High Quality images.
1. 轉载模型僅供學習與交流分享,其版權及最終解释權归原作者。
2. 模型原作者如需認領模型,請通過官方渠道联系海藝AI工作人員進行認證。我們致力於保護每一位創作者的權益。 點擊去認領
