LORA of Primav_style： A Low-Rank Adaptation of Large Language Models of myself

Before I turned 22, I was an illustrator with my own painting field and style. Here are my personal websites, Pixiv and Lofter. I used to earn a considerable income from painting and had over 20,000 followers. Now, I'm busy with scientific research and don't have time to paint anymore. So, I've trained a LoRA model based on my personal style. You can download it here and use it with large models.

Here are my original works.

illustration

Here are some of my painting and you can find more details in pixiv and lofter.

In the technical implementation, I carefully selected 100 of my personal paintings as the training dataset. I chose flux1s as the base model and used florensa to automatically annotate all images to generate training labels. The entire model training process was completed in the flux gym environment. Through this method, my personal painting style was successfully extracted and integrated into the LoRA model.

Here are some of the results of the model.

lora result lora result lora result lora result lora result lora result lora result lora result lora result lora result lora result lora result lora result lora result lora result lora result lora result lora result

Primav_style is an ongoing training.Only the painting style of faces has been trained here. The content related to bodies, a series of plants, etc., hasn't been incorporated into this model yet.

Year：2025.