You should expect PJ being flooded with AI art (control nets can even produce animation base frames in precise poses). You don't even need to setup a local model and multimodal AIs, like GPT4, have graphics parsing and generation builtin. I.e. they can generate art, parse it and process with LLM, which usually solves consistency issues, like 6 fingers (or 5 fingers in cartoon characters). Even the basic online generators are already better than most novice artists:

https://clipdrop.co/stable-doodle

So from now on it is about who uses AI more efficiently in their workflow.