Trained with "realistic", aesthetic Anime portrait images from Qwen.
Same training data. Low str around 0.5 works very good.
artyclaw-style-porper_krea2_v10.safetensors