V3 ("fix" version)

• Download the v3 file with the name "fix" at the end. Initial upload was the wrong one.

• Still a Lokr

• This version is much more potent. I would keep strengths between 0.5-1.0

• Didn't use any bypass or text refusal for my sample images only the Lokr itself.

• Recaptioned some images to improve sizes and concept adherence

• This version uses some training methods I haven't tried before so let me know if it works poorly

V2

• Overall Improvement. Multiconcept loras are hard...

• Size differentiation is better i think.

• Use with "realism" loras if you like. Should stack better than v1

• Recommended strength: 1.0-1.5

• Post images or give feedback!

• Doesn't require bypass loras if you just want busty women

:)

V1

Threw this together pretty quick. Krea is a bit quirky with multiple concepts, but this is just a fun v1

• I have better results using realism loras like Krea Realism v2 or Lenovo Ultrareal without any bypass filter

Can be used with the KreaFilterBypass set to 1.0 (increase for more NSFW)

Recommended Strength: 0.75-1.5

Issues: sizes can be unreliable (medium, large, huge, gigantic)