More About AI-generated Images

I have been continuing to experiment with the many AI applications that are generative. These apps generate new content. Generating text - such as with ChatGPT - has received most of the attention in the past year. Generating images came along later but has taken hold and now video and audio is the latest generative thing to do.

It is certainly tempting to create an "original" (Is that accurate?) image rather than pay for stock images or search for royalty-free ones that come close to what you want to use.

Below is one I made using Microsoft's bing.com/images/create/ which is just one of those available for free (at least for now). I gave a simple prompt: "man at keyboard using Bing to create an image of a woman using her smartphone." It is a good image. Certainly far better than anything I could do on my own.

I could revise the prompt and ask it to be in the style of van Gogh, or have him sitting outside or whatever. It's bit off from what I expected. By "image of a woman using her smartphone" I wanted her to have the phone in hand, not the man. I tried again using "man at keyboard creating an image of a woman using a smartphone" and one of the results is below.

Very different and closer to my intent. Here's an oddity - my revised prompt originally said "beautiful woman" and the app rejected the prompt as having language that goes against their policies. That's odd because it gave me four beautiful women anyway, including one that has a woman who looks like some comic book fantasy. (below)

For an article I was writing about hiking, I generated (I can't quite say I "created" it, can I?) several images. Very simple prompt: "man hiking in the forest in black and white." It generated 4 images and then I revised the prompt.

Not sure why he has a hiking/waking stick.

Adding color and a sunset

You can change the point of view - as with a ground-level low angle.

All of these are photo-realistic or painterly, but you are not limited to that, such as with this android blowing out its birthday cake candles.

It can become game of "stump the AI." Let's try "a woman and a robot walking hand in hand through a hall of mirrors."

I haven't always gotten exactly what I intended, but I have not stumped it yet.