Poelora (Pixiv Fanbox)

poeart

This is the lora I trained on my art a few months ago, free to use. Just drop it in your models/lora folder and add it to your prompts.

You'll want to reduce its influence to about 0.6; anything above tends to look deep-fried, and that's my bad for not knowing how to make loras correctly--unfortunately, unless you want to retrain the lora yourself on a faster computer than mine (toss all 2023's pictures into Kohya and you'll get roughly the same results), you'll have to work with that.

I trained this lora on top of NovelAI's model, so please use that or a model based upon it for best results.

If you're unfamiliar with Stable Diffusion, I recommend reading the beginner's guides linked on 4chan's /g/ /sdg/ threads, or watching instructional videos on Youtube to get started.

As a side-warning, Hires. Fix is a bit unpredictable with this lora (very likely my fault), and often you're better off sending a low-resolution image to img2img and applying UltimateSDUpscale or your secondary upscaler of choice. If you do that, however, make sure to use Adetailer to make the eyes more accurate before sending it off to the upscaler.