#2026-0825
昨天針對 Gemini 繪製 "合成圖" 的能力，做了三次 (即三組) 操作 / 實驗；詳情可參見
#2026-0824。
Yesterday, I conducted three experiments (or three rounds of testing) to explore Gemini’s ability to create “composite images.” For details, please see
#2026-0824.
Today, I selected another eight photos taken at different times and in different places and asked Gemini to combine them into more than ten “composite images” (合成圖). These images were not generated all at once. Gemini first created one image; after I gave it feedback and made some requests, it generated a second one. I then gave it new instructions or suggestions, and it created a third. The process continued in this way: whenever I made a request, Gemini responded by creating another image—the fourth, fifth, sixth, and so on—until all the composite images had been completed.
Yesterday, each experiment involved providing Gemini with eight photos and asking it to create one "composite image." Since I provided photos three times, I effectively conducted three separate experiments. Today was different. I provided only eight photos, but asked Gemini to combine them in a variety of ways, eventually producing more than ten composite images that differed greatly from one another in both form and content.
昨天，我每次提供八張照片，要求產出一張 "合成圖"；前後提供了三次照片，等於做了三回的實驗。今天有所不同，我僅僅提供八張照片，卻要求 Gemini 做不同的組合，總計產出十多張「形式」與「內容」都大異其趣的合成圖。
~ ~ ~ ~ ~
以下是我提供的 8 張照片 (Here are the 8 photos I used for the experiment)：
~ ~ ~ ~ ~
以下是 Gemini 根據我的照片，加上逐次給予的提示，所產出的 11 幅合成圖像 (Below are the 11 composite images produced by Gemini)：
= = =
相關文章 (See also)：
