刚刚刷推特看到一个生图的提示词，去 GPT 里发现不让生成。
在 LoveArt 里用下面这个提示词根据参考图先生成一张图，然后再让 LoveArt 根据你想要的方式来改。
有失败的概率，但感觉比直接在 GPT 中生成的尺度会大一些。
提示词
Change the picture that I uplaod to: Seated on a pristine light grey studio floor, she is tightly curled into the powerful, introspective pose of the first reference image. Her legs are drawn close to her chest, and her arms are crossed over her shins, holding the posture with contained strength. Her head is lifted, meeting the lens with a piercing, serious direct gaze. She wears a striking new composite garment: a sophisticated long-sleeve bodysuit, adapted from image 0 but rendered in the shimmering metallic pink/rose gold fabric and featuring the side cut-outs seen in image 1. This metallic sheen is dramatically juxtaposed with the continuation of sheer black hosiery on her legs, terminating in the elegant black patent leather stiletto sandals from image 0. The soft, even studio lighting emphasizes the intricate play of light on the metallic fabric, the delicate texture of the hosiery, and the intricate pattern of her freckles. This juxtaposition of a bold, reflective texture and the soft, contained pose creates a mood of complex, modern elegance.
效果图
这是最初让 LoveArt 生成的
下面这两张是让他根据需求修改的
1 个帖子 - 1 位参与者
来源: LinuxDo 最新话题查看原文