Theme: How to find a consistent personal style through training AI

As a visual designer, I don’t have a specifc style for my visual artwork.I always confused by how to explain my works and thread them together.Due to the functionality of lora I decided to train myself a model using my rendering and illustration.

However, the result was not that ideal at the end of stage.

base model i use:

a71a3b82a5c3090246660462887e0f0.png

The dataset:

3d1293d9ac0959eea99df2d12e0a263.png

training result;

image.png

image.png

ai12.jpg

I think 3 factors I would further developed at the next stage:

  1. cousicouslly choose the base model maybe the small one and it should fits to my style it self
  2. try long and accurate caption
  3. extend steps for training