|
I did not know there was the top of a head in the shot until many weeks later when the film was developed. There was nothing I could do short of painting the head out of the image. |
Many years later, September 2022 to be precise (so 36 years), I used DALL-E-2 todo some inpainting to remove the womans head. DALL-E-2 also does outpainting within the frame one is working in, here
it amazingly added the remainder of the lamp posts and trees within the square work area.
|
Next I used DALL-E-2 todo some outpainting to put in the missing front and heads of the horses, which I never had in the original, AMAZING!
|
In 2025 Google brought us NanoBanana which was very context aware and could do all of the above with a simple text of expand and show the whole carriage.
|
| Though all AI will have some aspects of this, I noticed an incredible similarity in images generated from totally different prompts, mostly of the background parts of the images. Below are 3 images with colour coded matching sections that clearly are related and the same even though at first glance they look different. This is most likely due to the noise/seed generator. As you flip between the images (by selecting the buttons) you will see even more correlations than the ones I have highlighted. |