R. Dorothy Wainwright || The Big O
Used kohya_ss gui.
Here are some parameters and helpful keywords:
• Checkpoint: AbyssOrangeMix3
• Trigger Word: dorothy
• Weight: 0.65 - 1 (0.75 is sweet spot)
Keywords:
Should typically work well with just the trigger word, but if there is trouble with specific details, try these keywords:
Face features: black eyes, red hair, black hairband
Outfit: black dress, white ascot with green gem
Mainly trained with her black dress, but a couple images included red / green, and model is versatile enough to do any outfit. It was difficult to obtain a good dataset for this character, mostly from anime screencaps, but I've had very good success using the LoRA with high detail checkpoints and may be able to train a V2 with higher quality images in the future.
If you enjoy the model, please share your images in the gallery~ <3
To produce similar results I recommend EasyNegative, this negative embedding was used in the sample images. I'm also using Clip Skip 2 and hires fix. VAE used is BerrysMix.