Krea 2

For this one, I've used a really small dataset (7 images) so it's overtrained, but it is really closer to the style I was looking for.

Showcase images :

• Turbo model

• 768x1152

• Euler / simple / 10 steps / CFG 1

Qwen2512

It was trained on 2512, but I prefer to use the original Qwen-Image model for inference.

Showcase images :

• Original Qwen-Image (BF16)

• Lightning lora (4 steps)

• 5 steps / 1 CFG

• Euler & Normal

• Lora strength 1

Klein 9B

Showcase images :

• Distilled model

• 768x1152

• Euler / normal / 4 steps / CFG 1

• Lora strength of 1

• When upscaled : x1.5 - 4x_NMKD-Siax_200k - Hi-res 4 steps - 0.35