Real AI imagegen workflows very often have more input from the human creating the image than a text prompt.

seems a little circular semantically. if it has significant human input its human generated moreso than ai generated, in which case we are saying the same thing.

Textures, sample images, something like that?

ControlNet offers a wide variety of additional input sources: https://stable-diffusion-art.com/controlnet/#Preprocessors_a...