<h2 id="heading-146">Imitation model. Modelled after Japanese 'catfight' wrestler and AV star Akari Niimura (新村あかり).</h2>Petite, cute, semi-muscular and gymastic-like, Akari got it all. Primarily debuted as a JAV star, she was later actually quite well-known for her performance in several catfight competitions, earning her the nickname "the sadistic queen" in both the catfight wrestling scene in JP as well as JAV. Perhaps attributable to her rather unique combination of power and feminity being palatable to a wide BDSM genre in JAV. <h3 id="heading-108">~~~~~~~~~~~~~~~~~~~~~~~~~~ Recommended LoRA strength: 0.9 (0.7-1.1 are plausible range dependent on checkpoints; higher values &gt;1 dont seem to do much..)</h3><h3 id="heading-109"></h3><h2 id="heading-110">Main Prompt:</h2><h3 id="heading-147">akariniimura_jav, breasts, long hair, brown eyes, jewelry</h3>variations: black eyes, tied hair, earringsfacial variations: parted lips, lipstick*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 this.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Checkpoints that work well for this LoRA:1) Realisian/PerfectWorld. Ok, depends on how you look at it: 1a) If you are into maximal facial resemblance and do not Mind a slightly more 2.5D render, then PerfectWorld is the model to go for. 1b) If you are still somehow obsessed with a more realistic representation and do not mind discounting some minor difference in terms of face-shape, Realisian is still a much better render. Basically, I can't decide which is better for everyone, so you choose. :P 2) RealisticMix. Similar to Realisian, but sometimes expression is slightly off.3) UltimateHentai Realistic or other photoRealistic Hentai models Works adequately well 0.9 LoRA weight, but prone to some bad/weird generations occasionally.4) MagicMix realistic. Facial wise, its ok. but you get slightly elongated body, and longer faces, which work against Akari's actual physical appearance. Still a "close-enough" render. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~Note for achieving good resemblance:This LoRA requires very minimal makeup prompt, since her features are pretty easy to define. If I need to break this down, alot of what defines Akari is her:1) Her eyes. Mainly because of her obsession with contact lenses, but thats built in, so do not need to worry bout it.2) Her parted lips expression. Which you dont have to address, since the LoRA take care of that directly.3) Lipstick. 4) a petite and semi-muscular built. If you want to, you can try using height prompts like "((short))" or "153cm" (yup, thats her height). In reality, does it make a difference? Well, not if she's not standing next to Michael Jordan or a tree..You can either use facial ADetailer directly with no prompt, or use a facial prompt like the one below if you want her with some makeup "face of akariniimura_jav, parted lips, smile, brown eyes, makeup, lipstick:0.3, &lt;lora:JAV akari_niimura v1_epoch_10:0.9&gt;" - denoising strength 0.45It 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:Nothing that needs immediate fixing, badHands, and occasional bad hand-body ratio. Just re-seed.in img2img upscaling, sometimes the shape of her face can be alittle off as most upscaler try to "beautify" her by shrinking the size of her pupils - I guess you can try running Adetailer again with very low denoise to try and fix that.Post-production edits/Reproducibility:Minor inpainting on some Badhands.Happy prompting. And do share [SFW] creations~

JAV Akari_Niimura

v1 Release. trained on 512x. ~ 86 images. But seems to do just fine. Might redo with 768 in the future with more data. Epoch #10 chosen.

Files with SHA256: 717b642867c03e280d82f190574447503a358efa62aa388f1846cde6488ac155