<h1>John Doe - A Style Lora </h1>This is more or less an experiment in creating Lora/Locon, and not meant to be a replacement or a strict recreation of John Doe's work. Please support the original artist! <a target="_blank" rel="ugc" href="https://www.patreon.com/JDArt">https://www.patreon.com/JDArt</a> This is a style Lora (a Locon, specifically. So you'll need the extension to properly use it) meant to emulate the style of John Doe (dimensional forms, dramatic lighting, "realistic-er" facial features). That's pretty much it.Seems to work best for me at around 0.9 strength, but feel free to play around with it. Lower settings will flatten things out and bring out more "anime" features (though certain tags like "smirk" tend to have a strong effect of bringing those out as well). It was trained on explicit images as well, so it should be pretty fine generating actual intercourse, and most subjects therein. Though you might have the occasional floating dick here and there.While the Lora also does well enough as you can expect at 512x512, I recommend using a slightly higher resolution to bring out facial features a bit more. Either by using a higher base res, or just turning on hires fix. It also seems to work well with character Lora as well as Lora depicting real people. I've given a few celebrity Loras a try, and while not a perfect recreation, their traits do show up, even if they were trained on different models and tagging systems. A surprise, but a welcome one!Since this was kind of a "lazy" Lora that I didn't meticulously curate, things might get a little off from time to time. Your mileage may vary, but the age-old adage reigns true, the better your prompts, the better your result. Still, it's absolutely possible to get decent results using basic prompting.Feel free to leave any comments you've got as well as any images you have made!The Lora is made using my usual workflow of training 2 separate Lora. One on full bodies, and another with head crops, and merging them together.

John Doe style

John Doe v1 - Initial version of the Locon. Good results at 0.9 strength. Prefers higher resolutions for better facial detail, but can generate at 512x512. I usually just have either the width or height at 512 and change the other to make different dimensions, and then highres fix after.

Files with SHA256: 7f79683b7cc987450e0ba1179420a5bdb236f7eb5b50dae7b88cd26be7c6b445