With keyword 'illustration' you get a fairly consistent style - black ink outline and detailed modeling with somewhat desaturated coloring. My personal preference is for somewhat succinct prompts that end with 'detailed illustration'.
With keyword 'retro Comic illustration' you can get some convincing old comic characters in a style similar to Kirby or Lee comics. Not part of the training dataset - it just lends well to such imagery.
Without any keyword you'll get generally more bold colors but more subtle ink drawing.
Flux will occasionally blur your background.
I'm using the model so far at just 1.0 strength. It plays very nicely with other style LoRAs
