Hi! This is my first LoRa, I hope you'll enjoy it! :)
It has been trained on Z-Image Base and works very well with Z-Image Turbo and Base!
All images showcased were generated with a weight between 0.8-1.0 with the res_multistep sampler.
It can generate characters, landscapes, anthropomorphic characters, fantasy style, modern style, sci-fi style, etc... It is quite versatile.
It can generate slightly NSFW content but it wasn't trained specifically on that, so you might want to add other ressources if that's your goal.
This LoRa was trained mainly on Makoto Shinkai anime style so the trigger word Makoto_Shinkai can produce better result.
Also, words like anime-style or anime aesthetic or any synonym is mandatory to trigger the LoRa.
Those images were generated without too much work put into it, you can certainly do better with custom workflows, better prompting, upscaling, etc...
Feel free to share your generations, your comments and drop a like if you enjoy this LoRa, so I can improve it and maybe create some more!
Thank you 🙏
Description
FAQ
Comments (4)
Somethings wrong here. Lora was ruining every generation.. very deformed. Turned lora off and photo looks just like the examples lol
Well that's weird, I tried it again today, this time using the Civitai AI creative tool instead of ComfyUi. I tried a basic prompt with a weight of 1.0 with the euler sampler and had no problem whatsoever (despite a bit less quality due to the euler sampler).
Can you share with me the details of your generations ? (Sampler, steps, weight, prompt if you still have it, UI used, etc...)
I'm curious about this issue and would love to investigate further ! Thank you !
my workflow was bad actually its all good now
@PrettyAIGirls Glad to hear it works now 🥳