If you like my work, you're welcome to leave a review. I don't really care, I do it for fun.
Lora Information
There's plenty of LoRAs for this character, but I had a private one I kept to myself a long while ago - I decided to refine it a little. A little is precisely that. The dataset is 80 images with moderate image work, the tagging was lazy low-effort auto-tagging with some manual adjustments. There is about.. 4 outfits.
See version details for advanced information (the kind of information LoRA Inspector gives ya).
I suggest weight of 0.8 ~ 1. Trained on Pony Diffusion V6 XL. Preview uses Prefect Pony XL v5.0.
Prompt Usage
The trigger word is maki_ozee, here are some sample prompts. The firefighter coat, witch-looking hat didn't have enough data, but they are there (coat, witch hat).
The following are examples, use anime style tags, use quality tags, use expressions, angles, backgrounds. Gallery for prompt guidance. I forgot to use score tags.
Orange Jumpsuit
maki_ozee, 1girl, solo, purple eyes, black hair, bangs, long hair, ponytail, pants, orange pants, jumpsuit, cowboy shotOrange Pants + Tank Top
maki_ozee, 1girl, solo, purple eyes, black hair, bangs, long hair, ponytail, pants, orange pants, clothes around waist, tank top, black tank top, cowboy shotDenim Jeans + White Shirt/Pink Shirt
maki_ozee, 1girl, solo, purple eyes, black hair, bangs, long hair, ponytail, shirt, white shirt/pink shirt, short sleeves, belt, denim, jeans, cowboy shotMilitary Outfit
maki_ozee, 1girl, solo, purple eyes, black hair, bangs, long hair, ponytail, hat, garrison cap, uniform, military uniform, necktie, skirt, pencil skirt, armband, cowboy shotAnd that's all I really have to say for now. Later!
Description
See description for trigger word/prompt guidance.
Network Rank/Alpha: 32/16
Optimizer: Adafactor, cosine with restarts 3
Learning Rate: default, 0.0005 and 0.00005
Others: Noise Offset of 0.05, Multires Noise Iterations of 6, Discount 0.3, SNR Gamma of 5
Dataset: 80 images, 8 epochs, 6 repeats, 4 batch size, 960 steps + Keep Tokens, No Flip Augmentation, Shuffle Captions, No Regularization
















