SDXL_anime_clear_vae
This is a VAE (Variational Autoencoder) for the SDXL animation model.
Compare to SDXL_Base_0.9_vae (Original)
SDXL_anime_clear_vae | SDXL_Base_0.9_vae (Original)
Base Model: Animagine-XL_4.0_Opt_clear
It features clearer rendering compared to the original, and is characterized by a deeper color expression, primarily in Reds.
However, in fine detail, there is a slight tendency for detail to be subtly crushed or lost.
Tips
License
Description
FAQ
Comments (5)
Is there a trainer to train the SDXL VAE on 1024x1024 images with no OOM?
I'm not performing any additional training (fine-tuning); I am only applying scaling to adjust the outputs of each layer of the model.
qekNov 30, 2025
@easygoing0114 Yes, it's called merging, not fine-tuning
noteも見てます。
色々なVAE試してみましたが、色合いがベストでした！
きりっとした色合い、ビビットすぎず。
自分のイラストにピッタリです。
コメントありがとうございます！
どんどん使い倒してください