Flux 1.0 dev:
This is a retrained version of Julie Newmar's Catwoman, built on Flux 1.0 dev. I've completely remastered the model from scratch, enhancing all images with SUPIR to add texture and details that were missing from the original dataset. The result is a good balance between resemblance, quality, and flexibility.
For best results, use the default Flux workflow with either euler or deis samplers, set between 20 and 50 steps. Keep the LoRa weight at 1 for optimal outputs.
This model is part of my ongoing research and learning process, so I may release an improved version in the future. Feel free to experiment and please share your creations!
Hello! This is my first LoRa ever after many attempts. It is intented to honor one of my childhood heroes: Julie Newmar Catwoman from the 1966 series. I know it's not perfect, I made it with screencaps of a 720p Blu-Ray, it's the best I could find.
Context:
Julie Newmar is an American actress renowned for her iconic portrayal of Catwoman in the 1960s Batman television series. Her portrayal of the feline-inspired villainess left an indelible mark on popular culture, characterized by her sultry charm, playful demeanor, and sleek black catsuit. Newmar's Catwoman became a quintessential figure in the show, captivating audiences with her alluring and enigmatic presence as she engaged in witty banter and captivating schemes opposite Adam West's Batman. Her performance as Catwoman continues to be celebrated as a timeless representation of the character in the realm of classic television.
SDXL:
Usage:
Activation Words: catwoman woman
If you don't want to render her with the mask just type it in the negative prompt.
V1
After many test I recommend using these settings for a better output.
LoRa Weight: 0.6
Clip Skip: 1
Sampler: DPM++ SDE Karras
CFG Scale: 6
Resolution: at least 1024 x 1024 px
Don’t use ADetailer, it doesn’t blend very well, same with Restore faces. If the face is too far you can either Hires. fix or inpaint if you have enough time.
Hires. fix for optimum quality
Denoising strength: 0,35
Upscaler: 4xNMKD-Superscale-SP_178000_G or 4xNMKD-Superscale-_150000_G
V2
Same as V1 but the LoRa weight is 0.7
Have fun and if you can, share your creations here! As I said, it's my first LoRa so any suggestion will be very much appreciated. Thanks!
Description
This is a retrained version of Julie Newmar's Catwoman, built on Flux 1.0 dev. I've completely remastered the model from scratch, enhancing all images with SUPIR to add texture and details that were missing from the original dataset. The result is a good balance between resemblance, quality, and flexibility.
For best results, use the default Flux workflow with either euler or deis samplers, set between 20 and 50 steps. Keep the LoRa weight at 1 for optimal outputs.
This model is part of my ongoing research and learning process, so I may release an improved version in the future. Feel free to experiment and please share your creations!