I uploaded a V2 version, but i'm not sure it is better in every way. It creates better camouflage pattern even without prompting and fewer nonsensical rank signs. It seems to struggle more with complicated anatomical composition. Likely there will be a (final?) V3 soon. Back the the plot:

The official modern german BDU ("Kampfanzug" literally translates to BDU) . Why? Because "german soldier" doesn't HAVE to give you that WWII look, all the models seem to produce.

Use trigger "!bw" Add possibly "Flecktarn Uniform" and "combat boots", depending on your image, weight and workflow. It should add reliably as headgear the (official) "Flecktarn cap" or "beret", if you prompt it. No helmets trained though. Also nothing past the actual BDU. No heavy jackets weather gear and so on.

If the pattern is too washed out " the camouflage pattern is rendered in high detail and contrast, full color" fixes that. I found without "Full color" that "high contrast" prompt makes Flux go black and white?!

The Lora was a originally a quick first try, but it works actually quite well. Except for nonsensical rank signs.

You should use !bw to trigger the image generation.

trained but no official trigger: Flecktarn cap or (black or red) Beret for headgear to trigger the image generation.

Use wears combat boots if barefoot. to trigger the image generation.

Weights for this model are available in Safetensors format.

Download them in the Files & versions tab.