<h2 id="ww1-french-soldier-uniform-3956ldrff">WW1 - French Soldier uniform</h2><h3 id="how-to-use-flux-version-47in1muo0">How to use FLUX version </h3>You have to use flux way: use verbose prompts and use words from the caption. You can do any style, woman soldier, anime, illustration, whatever. Just prompt wisely. Lora strength to 1 seems mostly fine but you have to experiment obviously, as well with flux guidance. Top 20 Nouns and Adjectives by Frequency: French: 28 blue: 28 WW: 20 soldier: 20 leather: 18 military: 17 helmet: 16 rifle: 15 uniform: 13 shoulder: 13 soldiers: 13 black: 12 attire: 11 rifles: 10 brown: 8 uniforms: 8 pouches: 8 white: 7 boots: 7 right: 7<h3 id="how-to-use-gn05h8kqx"> How to use</h3>This model has been trained on base SD 1.5. It should be compatible with almost any checkpoint although quality may vary. It works with male or female characters. There is no keyword to invoke but I personally use "WW1 French soldier, a man". The LoRA weight depends heavily on what you want to do so it's hard to recommend something specific, I would say from 0.5 to 0.9 give the best results in normal scenarios.<h3 id="historical-accuracy-51ghp5xt8">Historical accuracy</h3>This LoRA imitates the common uniform of French soldiers during WW1 from around 1915 and up to the end of the war (Bleu Horizon). It does not feature the uniform from the start of the war (red pants, cap, no helmet).<h3 id="dataset-vsa9dkz6b">Dataset</h3>Only images of costumes or reenactments were used because it was far easier than using original images which are mostly black and white and in bad quality. An alternative would have been to use images from WW1 movies, but costumes were faster to acquire, as only 16 images were needed. I plan to make other LoRA themed about World War 1 so don't hesitate to give advice and feedback!

Files with SHA256: 6c4ad358dd3fb1a487ebcf72c89819bf7a57d8814a83002753893d8b7ae50e67