随手涂鸦两笔让 GPT Image 2.5 出四张图
作者用 ChatGPT 里的 Sketch 画三秒乱线，一句话让模型朝四个方向创作，展示涂鸦转成时装、建筑、狐狸和星系。
GPT Image 2.5 doesn't need your talent anymore. It needs two lines. I wasn't even trying. Opened @sketch in ChatGPT, dragged my mouse into a random squiggle three seconds, no idea, no composition, no meaning and typed one line: "turn this into the most creative image you can, in 4 completely different directions." 15 seconds later, on my screen: - a couture campaign where my line became the dress - a futuristic building on the water at sunset - a golden fox under a starlit sky - a galaxy woven out of the same two lines No edits. No designer. No brief. All from one scribble I'd be embarrassed to show a friend. You don't get paid for skill anymore. You get paid for coming up with the first two lines. Sketch is web-only right now go test it before everyone floods in.
这条是引用着另一条写的：x.com/openai/status/2097394956457623964