"One Image Is All You Need" Example LoRAs (Pony Edition)

Details

Download Files

About this version

A man and a woman dancing.

Pony retrain of /model/206187?modelVersionId=235112 Uses regularization data.

This was trained with long captioning so all of "ballroom dancing, from side, man, woman" are pretty much mandatory to get an effect like the original image. "score_9, score_8_up, score_7_up" also help it look more like the original image's style, because otherwise the regularization data shines through. It's highly recommended that you specify clothing as the man in particular tends to be mostly nude without giving it hints about what he should wear.

Full captioning was "ballroom dancing, from side, woman, pink dress, choker, dark brown hair bun, blush, back, man, white coat, short black hair, black tie, black pants, belt, theater stage background, bokeh" so all of these will capture some aspect of the original.

Works best at medium strengths, or varying the weight from low early on to higher later.

Model description

These LoRAs are associated with the article One Image Is All You Need. They all use a single image dataset. Go read the article for more information on how to use these properly.

@ZaphodScotsman requested that I retrain /model/206187?modelVersionId=235112 on Pony, and the results were better than I was expecting. So, here it is.

I didn't initially retrain these on SDXL/Pony because anyone can do it with the training data I provided. I did check they worked but didn't release it.

In some cases, I used my standard regularization training data to improve results. The pure single image LoRAs are marked with "noreg". See each LoRA's details for more specific info on proper use.

Also different is that I use Prodigy as detailed in this classic article: https://civitai.com/articles/3105/essential-to-advanced-guide-to-training-a-lora

Images made by this model

Sort by

No Images Found.

Model Type	LORA
Base Model	Pony
Published	9/23/2024
Trained Words	ballroom dancing, from side, man, woman