A fun LoRA trained on images from /r/StillLifeDoodles/
suggested strength is 1.0 for model, and 0.5 for clip, but i didn't test a lot of variations.
It's not very flexible but with multiple tries it will probably work.
Some prompts that worked for me, and I suggest you try similar short prompts:
Photo doodle, watermelon crying
photo doodle, big mac with hands
photo doodle, watermelon in a basket
photo doodle, angry piece of cake
photo doodle, a cool dog
photo doodle, two apples
(cool works for variety of subjects)