AI image making

So I come upon an article on Medium about creating images from text descriptions, and it occurs to me this might be a cool way to make book covers. The technologies the article suggested required complex signups, so I searched and found NightCafe instead and tried it.

I chose “oil painting” and the “coherent” filter, typed in Victorian street Grimshaw and got this:

Apart from the weird sky, I cannot tell what the hell that figure is supposed to be. I can only tell it’s a person because it has feet.

Next effort:empty victorian street streetlamp cobbles cabriolet cloudy day.

Huh. Skies full of cobbles. OK, how about Victorian London, the artistic filter, and the “steampunk” setting:


The architecture is better, but what is that thing? I write historical mysteries. OK, so no steampunk setting, just Victorian London street and “oil painting”, back to the coherent filter:

Another creepy non-person. And that vehicle? Yeah, right.

OK, so historical isn’t happening. I’ll try what it’s supposed to do, fantasy. Moonlit Thames river boat, with the “heavenly” style and the artistic filter.

So, love the idea. Not thrilled with the execution. But I’m interested in looking more into this for sure!

2 thoughts to “AI image making”

    1. Sometimes “steampunk” is fun; other times just, yeah, Lovecraft-y.

Comments are closed.