Trained on more images with captions less likely to overly influence generation.
Main trigger: "slippery"
Optional triggers: "fall" "rear view" "on liquid" "slipping on liquid"
