跳到主要内容
Most people write one long paragraph and hope the model weighs it correctly.
It does not. Ask for a character, a location, a mood and a camera move in a single block of text, and the model quietly decides which parts matter.
On the canvas you split it.
A subject node for who