(This lora is relatively old/outdated)
ありがとう 🍌
READ VERSION DESCRIPTIONS.
Description
But what has changed??
This version uses the same dataset however, is trained upon anime final prune (NAI). There is much less keyword wrangling involved and his suit should appear more accurately with less prompting.
If you're still having trouble, feel free to add gakuran to prompt
FAQ
Comments (3)
Looks awesome, but what's anime final pruned nai and why is the training better on it ?
Anime final pruned is the leaked Novelai checkpoint. it is an anime focused model based on SD 1.5. Training on it gives better results when making LORA of characters VS merged checkpoints from what I was told. This could be due to merges (for example, training on anything v3 orange mix etc.) creating unwanted results
Most anime checkpoints on here ARE based on either Stable Diffusion 1.5 or Anime final pruned so the character should appear mostly consistent no matter what derivatives you use (I used Anything V 4.5 in the preview images)
@randombanana awesome i didn't know that, i used to train on anything 4.5. thanks