Generate art images using a Diffusion Transformer (DiT) model
S: Small (fastest), B: Base (balanced), L: Large (best quality but slowest)
How many images to generate (1-16)
Set for reproducible results