Extended fanboying about Jim Aparo in the OP for my original Illustrious version, if you want to know more. Samples above are straight gens with no detailers or upscalers, using the model it sez on the pic – except for the splash image, which was generated in Base and upscaled in Turbo as a proof-of-concept.
I had just about finished re-tagging my Jim Aparo dataset for Chroma (meaning both tags and NL) when Z-Image Not-Turbo came out, so I figured, why not?
As advertised, Z-Image base is sloppier and slower than Turbo but has more room for creative freedom and prompt adhesion. Less as-advertised but more as-expected, Base LORAs work on Turbo, but you have to give them a rather heavy push.
As tagged in the images themselves, you have to run Base LORAs on ZIT at 1.5 STR to match the effect of 0.8 STR on Base. Note that Base loves to oversaturate – I’m prompting for lower saturation and paper grain texture to offset the lurid colours. Prompts and workflow are encoded in the sample pics.
The LORA was trained with a mix of natural language and Booru tagging, including jimaparo and aparobats keywords, but I find they’re not necessary – “A comic book illustration in the style of Jim Aparo” will generally do the trick. You might need to specify blue costume and yellow bat-symbol for the bat-dude if that’s what you’re trying to achieve, though.