Based on portrait images in a cinematic style.
This model is surprisingly good at lending a more natural, photographic look to images without imposing an overly strong effect or messing up prompt adhesion / cohesion.
It does however seem to have a "bug" where on some prompts vertical banding appears.
Trigger words are "In the style of ohwx" but works well without too.
Description
Trained with "In the style of ohwx" as trigger, but may work without.
FAQ
Comments (8)
This is insane. I'm not sure how you did it, but this is probably one of the most realistic generators I've ever seen. I couldn't tell you about 50% of these were AI gen'd if my life depended on it.
Thanks! To be honest I think at least in part I got a really lucky training run. This was supposed to be an experiment training run that turned out really nice.
U are the best, the pix differ from KI to Reality, i love you
Really great Flux Lora, I highly recommend.
EDIT: I did notice it is causing the "Flux lines" in some of my images but depends on a lot of things.
happened same to other lora called 'openflux-fast-lora' even at 0.25 weight, shames as it was speed and concept changer
@lokalfokal You can try using it with a Lora Loader (Block Weight) node and setting layers 1-3 to zero
https://www.reddit.com/r/StableDiffusion/comments/1g7b76e/a_possible_fix_for_the_gridlines_effect_when/
@J1B nice it worked crystal clear, i just had to use custom loraloader by KJnodes to bypassed tuple. Thanku o7
This lora makes a big difference to the skin. However, if another Lora is used that brings a character with jewelry into play, ear rings and necklaces disappear.
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.