Louis de Funès, 87 images, 2850 steps rank 16 alpha 8, trained on Flux 1.dev with fp8 T5x clip, learning rate 0.0005 (or maybe one more or one less zero), trained on ComfyUI with a workflow found on civitai.

I am not sure the trigger word defunex is actually so important, I always put Louis de Funes in the prompt usually, so just in case, add Louis de Funes or de Funes in the prompt...

since the training images weren't super sharp, as a result there is some kind of a base blur to the generated images, not too crazy but it's there, probably some technique and upscaling can reduce this, but I didn't tried so hard in the example images