New LoRa Work in Progress for Z-Image Base

Another early preview of an image using a new LoKr trained on Z-Image Base from the same dataset as my AfterDark LoRA for Z-Image Turbo. Not bad, it helps with lighting and shadows quite a bit. Though Z-Image Base can be very sharp to begin with so ensuring that isn't taken away is difficult. Z-Image Base seems to almost over sharpen images and it also adds better depth of field. That actually helps address some of the things my LoRA was addressing with Z-Image Turbo's shortcomings which is great...but it still falls short on lighting. Hair looks a lot better in the base model as do clothing textures. So if I can just get the other lighting and photographic qualities from my LoRA into a new one for Z-Image Base and not lose out on the increased quality from the base model, I'll be happy to share it with others.

I've attached the same image generated from the base model without the LoKr. In this case the image using the LoKr is sharper with better detail (though this isn't always the case). However the lighting is dramatically different and it helps illustrate what I'm after and where things are at currently.