
Imitation model. Modelled after veteran Japanese AV actress Yuuri Oshikawa (推川ゆうり).
By now, Yuuri probably worked in the AV industry close to 11 years. LIke many other AV actresses, she begin as a gravure model, but slowly gained popularity with NSFW photoshoots. Those of you that follow the AV-turned-idol group Ebisu Muscat, might actually know her as well.
Yuuri's feature is well regarded to be slightly chubby, but with body fat around where it really matters: curvy, feminine and . Its to no surprise, she is a killer with the NTR genres (even back when she debuted), desperate hosuewives, late-at-work, business trip affairs etc. But whats distinctive for Yuuri, is probably her slightly wavy bob-cut (short hair), juicy lips and the baby fat around her cheek. The LoRA captures these pretty well.
~~~~~~~~~~~~~~~~~~~~~~~~~~
Recommended LoRA strength: 0.8 (0.6-1.0 are plausible range dependent on checkpoints; higher values >1 exaggerates expression.. 0.9 can be good in certain scenarios)
Main Prompt:
1girl, mature woman, yuurioshikawa_jav, breasts, short hair, brown eyes,
variations: bob cut, wet hair
facial variations: parted lips, lipstick, pink lipstick, makeup, thick makeup, teeth, smile, grin
*Use of facial prompts depend on checkpoints, you can make a case for using it or not - personal preference
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Checkpoints thatworks well
1) No strong favourite. Most Asian-realistic models performed quite adequately, with varying degrees of physique resemblance and bias. I personally love Realisian's render, but you can try other checkpoints.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Note for achieving good resemblance:
"Mature woman" can be useful to add some aging, since most SD checkpoints tend to create females that are much younger. Using "milf" is risky, and can tend images to being NSFW. You can technically do without a separate prompt for face Adetailer. But, if you want better accuracy of her facial features, you can do the following:
How to maximise facial resemblance:
1. Generate the main prompt using a LoRA strength of 0.8-1.0. so its probably safe to say that this weight is stable for use.
Note: In this step do not add makeup prompts, if you want to retain flexibility and prevent overbleeding. I find it OK to just have makeup prompts like "lipstick" used in the Adetailer prompt.
2. Next turn on facial Adetailer, and put something like
"face of yuurioshikawa_jav, short hair, lipstick, makeup, parted lips, brown eyes, <lora here>"
For the loRA weight in the Adetailer, I recommend 0.8. Experiment this, its different for every checkpoint.
For Adetailer strength, you can try 0.4 - 0.45 for close-up or cowboy shots, and even 0.5 for full body ones.
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 0.15 -0.20 (anything higher will overcorrect face during the process), and turn off Adetailer. Do not use anything more than 0.25 else the eyes might look bit weird.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Issues:
No issues so far, apart from some badHands (in some checkpoints).
Post-production edits/Reproducibility:
Minimal badHand edits. all img2img original upscales.
Happy prompting. And do share [SFW] creations~
描述:
v1 release. Trained on 300+ images.
Epoch 12 chosen out of best multi-angle consistency. (superior support for landscape, vs other epoches).
训练词语: yuurioshikawa_jav,mature woman,breasts,short hair,brown eyes
名称: 537142_training_data.zip
大小 (KB): 148999
类型: Training Data
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
名称: JAV_yuuri_oshikawa_v1.safetensors
大小 (KB): 36999
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success