This is my attempt to forestall becoming a uninformed Luddite with regard to artificial intelligence (AI) and its positives and negatives when using existing images instead of text to generate new images. I’ve decide to explore some of the software options. Many free programs are available, though free is a fluid term. To produce higher quality, higher resolution, and sharper images, many programs charge credit based “fees” or require subscriptions. I plan on using the same image of my nephew and grandniece as the starting image, if possible. If only text to image generation is available, I will use this prompt: Young bearded man wearing a blue shirt sitting outside on a porch holding his baby daughter. 

Dream by WOMBO uses text and image prompts in it AI image generation. With the optional image prompts, though, there must be at least one descriptive word in the text prompt box. I used all three prompt options to produce the results below. Dream has a free level which provided one image for each image generation request and has ads. Monthly, annual or lifetime premium plans are ad free.

Image and One Word Text Prompt

The optional Input Image is used as a reference for the final output. I was surprised at the diverse images generated. The only difference was the text prompt word. The colors reflect, though don’t necessarily duplicate, the original photo

Image and Full Text Prompt

Text prompt “Young bearded man wearing a blue shirt sitting outside on a porch holding his baby daughter” and the original input image. I picked a style for each new image created. Some “create” actions produced a mishmash of strange lines, colors and body parts. One action produced an NSFW Content warning, but I don’t know why because only the style selected had changed, not the text prompt.

Text Prompt

Text prompt Young bearded man wearing a blue shirt sitting outside on a porch holding his baby daughter. I selected different styles with the text. I was surprised that in some images, the child was missing. I love the details, such as the earring in the first image. The baby looks a bit possessed because the light reflections are in the whites of the eyes. And the father’s hand on the child’s back is child-sized.

Edit Image (BETA): Edit with Text

You can describe a change you would like to see in an image you upload. I am not sure which image formats it will accept. It felt strange to use words to tell a program home to modify a face. In the Edit with Text box, I entered “make beard on chin longer.” A thick dark beard was added. To me, Jeff’s entire face now looks more AI generated than realistic. It doesn’t feel real to me.

I learned a hard lesson about using a program I don’t know well. While I generated lots of images, I didn’t save them as finalized. So, when I closed the program and went back to review the styles I had used, the images were no longer there. Luckily I had downloaded the ones I wanted to used. Some of the style captions may be incorrect. New images are saved and downloaded as .jpg files. If there are any any FAQs, guidelines, or manuals, they are well hidden. Maybe you need to be a premium member to access. The free version is basic and generally intuitive.