The LoRA has strong influence over the details of the image. It tends to make subjects younger. Women go from late 30's to early 20's.

It seems to make fashion and object changes. It seems to be heavily influenced by late 90s or early 2000's.

With all those big changes and loss of control for the user... I can't really say it makes the image seem more real, candid or amateurish.

The Krea2 LoRA works incredibly well, especially with prompts like

"A amateur photo of x"

or

"Photo shot on an iPhone."

Increasing the strength up to 2.0 improves a lot of NSFW generations, at the cost of possibly some artifacts in hair structure. Only noticed it on blonde "thinner" hair, but noticed that throughout LoRA's in general for Krea2, including my own trained ones. Might just be a Krea2 gimmick with multiple/stronger lora's that don't stack well when they affect the same vectors/layers.

Can't verfiy what someone else said about subjects appearing younger, even prompts like:

"A amateur photo of a blonde woman sitting on a table in a cafe. The picture is a close-up of her upper body and face. She is smiling while holding a baguette. Photo shot on an iPhone."

Will prompt a mid 30 woman with slight creases at the eyes. So no bias towards younger women from any of my testings.

I think the lora completely excels at providing the wanted "amateur" look without impacting any character loras.

I always train 2-3 people LoRA's whenever a new base model comes out and with Z-IMG (both turbo and base) it was nearly impossible to have more than 1 lora because every style lora unwillingly changed the depiction of the person.
This lora for Krea2 changes nothing about my character loras while perfectly shifting the image towards more of a candid-picture style - thanks a lot! 🫶🏽

Paid 5k buzz for this amazing lora. words can't describe how grateful i am for a krea 2 version! thank you so much for creating my fav lora from z image turbo to krea 2! this lora is amazing!

Your lora is the most effective among all realistic loras, especially for those with insufficiently trained characters. Using an intensity of 1.5 to 2.0 can significantly enhance the fidelity and realism of the character lora

Drastically increases the quality of renders, so to me is the best LoRA for realism so far.

If you're not a god, you're the closest thing to it we've seen yet. This krea 2 realism is absolutely insane, masterpiece work!

just recently began to discover for myself wonders of KREA 2 model. thank you for your contribution, very promising LoRA to my modest opinion, GG's

Does anyone have any workflow recommendations to go along with this lora? :)

Awesome! Is this version going to be optimized for Krea 2 RAW or Krea 2 Turbo?

For someone who would want to train something similar, would this fall into Character or Concept? I assume concept.

Is this trained based on the Krea Raw model or the Turbo model?

Good woman aesthethic with this lora and irl style, downside that i do not like is that its framing images very close, too close

your Krea2 v0.5 is actually very promising, good work !

Loving this lora alot. My only issue is that it's hard to generate below average/ ugly women with it, other than that 10/10 lora ⭐️

Can this LoRA be combined with my own Character LoRA?

For some reason, on krea 2 turbo it creates an image in the style of a 3D render, not realism.