v2 - More stable and more flexible so far, use a v1 dataset combined with new data; you should generate without any problems.
To change the outfit, simply add (alternate costume), then the alternate outfit you want to use.
------------------------------------------------------------------------------------------------------------
v1 - Initial project status. Preparing a new version with more accurate data coming soo
Errors found:
If the prompt is not specified correctly, it will probably generate multiple views and add more people.
The raccoon usually appears 😅, Use the negative prompt.
LoRA may not be very flexible with character positions, this because it was only trained with a minimum of 10 images with data augmentation.
🔨 I'm already fine-tuning the dataset to publish a new versión.