CrazyWhatever – Subtle Chaos, Creative Control
CrazyWhatever is a playful and versatile LoRA that introduces small but impactful changes to your images. It enhances or changes sharpness, color intensity, and detail in unique ways, often adding a touch of texture or unexpected visual twist without overpowering the original prompt.
It’s especially useful for creative experiments, working well in both realistic and stylized compositions. While it’s not a full-blown style LoRA, it’s a great companion to other models, adding nuance and character through subtle visual shifts.
Note: CrazyWhatever isn’t meant to do everything. Some of its effects are minor or depend on specific prompt combinations. Still, it brings a wide range of creative possibilities and helps make your outputs feel more dynamic and alive.
SFW/NSFW
Flux can easily go up to 2.0
SDXL V2 is mor efor NSFW and even a bit crazier
IL go about 0.3-0.5 most of the time
Anything you find of mine outside of Civitai is shared without my permission.
Please do not reupload my files. Just link people to my Civitai profile instead. Thanks :)
https://civarchive.com/user/tehalex86
Description
Well IL was still missing
Go between 0.3-0.5 . depends on how wild the prompt ist
Or well...go full crazy ^^
Anything you find of mine outside of Civitai is shared without my permission.
Please do not reupload my files. Just link people to my Civitai profile instead. Thanks :)
https://civitai.com/user/tehalex86
FAQ
Comments (8)
I'll have to download the new SDXL version. I do love the Flux one!
Thank you. Hope it's doint well. Did just some tests, but V1 is more the refiner, while V2 is a bit more crazy ;)
Could you allow for lower values than 0.5 (for the SDXL version) in the on-site generator. I feel that 0.5 might be too strong in some cases.
Sure thing. You can go as low as 0.1 now
@tehalex86 I have to say: I've been "avoiding" (not really, just: too lazy to try new things) this LoRA for far too long... it's actually quite impressive what it can do (at least in my limited experience with it).
@fbahr Huge thanks for the kind words. Was a fun experience/test with this one. You can get lucky sometimes with it. Was a wild training session
A very wonderful and beautiful LoRa 😁
Not sure yet, but thank you <3 . So much to learn with Z-Image training. So many fails the bigger the data set. But this was the first version I was somehow happy with