Home Artists Posts Import Register

Downloads

Content

This one is different from the usual LoRAs. For starters this is trained on real pictures, the original idea was to train this over different clothes and undergarments to have a wider spectrum of clothes in the models. This failed horribly, as the model tends to change the overall picture to a more realistic one.

To use it while also keeping the anime style, you'll need to lower the weight (usually 0.3 works). Once you get a good starting picture, go to the img2img tab, and remove the LoRA from the prompt, this should fix the style of the picture.

As this is trained over realistic pictures, it should work also on your realistic checkpoints.

Files

Comments

No comments found for this post.