My first LoRA trained for Krea 2, v1.1. (first one uploaded on my test account)
A character LoRA created from a small, varied dataset, focused primarily on maintaining a consistent facial identity while allowing different poses, outfits, hairstyles and environments.
Dataset created using OpenAI, upscaled and enhanced on OpenAI, original character with facial features created with custom parameters in details on OpenAI, inspired by Balkan/Eastern Europe beauty standards.

Trigger word: zvynara
Recommended LoRA strength: 0.7–1.0
Base model: Krea 2

First version — more testing and updates coming soon.