Kittew style for Z-Image Turbo

No special treatment or trigger word needed. Z-Image ignores trigger words either way, at least that's mine experience. Just do what you like and you will gain semisuccesful results.

So it has been a couple days since Z-Image has released and honeslty? I am pretty excited.

I want to believe this is the new era of local AI generation. Things were getting a bit stale with SDXL, it's still good due to years of community support, but we were hitting our limits with it. Flux is an unoptimized mess with horrible weights and I was feeling kinda sad about the whole thing.

And here, Z-Image comes. Not only is it ON PAR with Big Tech image models, it fits on 11 GB! If using a Q8 GGUF - 7 GB, if using FP8 - 5 GB. This is a miracle and something I think most of us thought to be impossible. But here we are.

This Lora is pretty raw. Nobody yet knows how to make good LoRAs for Z-Image, no videos, no guides, it's time we figure this out on our own to achieve the best results. I am also going to put as much work as I can towards making better and better LoRAs for Z-Image so that it can become the succesessor of SDXL we all deserve.