微軟 Bing Image Creator Dall.E 作畫紀錄2 神經狀的發光體

有時候做完功課，或禪修的時候，閉上眼睛會看到的東西，只會出現的幾分鐘。它會先有個輪廓，然後由暗慢慢轉亮，亮到最大後，就會慢慢轉暗，然後回歸虛無。所以想藉著Dall.E，看能不能畫出看到的東西。

圖一

給予的描述是這樣：「全黑的背景，中間有個像核桃的，上下較扁，左右較寬的金黃色亮光，金黃色亮光佔了整個圖片的80％，亮光的左邊輪廓有點缺陷，亮光中間有個小黑洞，亮光的光像密集的神經網路，由中間的黑洞向外放射，布滿整個光體。相片風格」。

翻譯出來的指令是這樣：Create an image with a completely black background. In the center of the image, there is a golden-yellow glowing entity that resembles a flattened walnut with wider horizontal dimensions. The glowing entity takes up 80% of the entire image and has a slightly imperfect contour on its left side. There is a small black hole in the center of the glowing entity, and the light emanating from the entity takes the form of a dense neural network radiating outward from the black hole, covering the entire entity. The overall style of the image should be photographic

畫出來是這個樣子：

圖二

但是光線沒有布滿整個光體，所以改成

給予的描述是這樣：「全黑的背景，中間有個像核桃的，上下較扁，左右較寬的金黃色亮光，金黃色亮光佔了整個圖片的80％，亮光的左邊輪廓有點缺陷，亮光中間有個黑洞，亮光的光線像密集的神經網路充滿整個光體。相片風格」。

翻譯出來的指令是這樣：Generate an image with a full black background. In the center, there is a flattened, wide, walnut-like shape with a golden yellow light that covers 80% of the image. The left edge of the light has some flaws in its contour. There is a black hole in the middle of the light, and the light’s rays resemble dense neural networks that radiate outward from the center. The entire light body is filled with these neural network-like rays. Please use a photographic style to generate the image.

畫出來是這個樣子：好吧，黑洞變大了

跟第一張沒差多少，這大約是眼中圖像剛開始出現的樣子，但是像描述那樣的最終狀態， AI一直生成不出來，甚至越來越抽象，最後我放棄了。看來指令可能還有得研究。