Ebora Artstyle Lora trained with 512 images (40-60% of the dataset is high quality images)
Removed chibi, doodles, English commentaries, texts, and background noises that were going to disrupt Lora's training if left untouched.

There are only two images with English commentary, and I made sure to put tags for them.

Autotagged first with wd-vit-large-tagger-v3 / wd-eva02-large-tagger-v3 taggers and personally reviewed all the dataset tags one by one to see if there are any imaginary tags or missing tags.

There may be some images where I missed tagging some elements, but not every Lora is perfectly tagged and over tagging can be poisonous some time to time.

This is a moderately to highly curated dataset, starting from approximately 1300 images and ultimately reducing to 512 images, personally edited by me with meticulous attention to detail which was horribly time inefficient.

There are some more things I've done, but I forgot the things I did to the dataset.

Recommended Strength : 0.7 - 1
The dataset is updated til July 17 drawings

The artist : https://x.com/youngmanisdown