My friend Lou asked me to use my AI skills to make an image he wanted for a joke email. Summarizing his description, in its final form the prompt was basically
As Jesus enters Jerusalem he is practicing his circus bareback riding act. He is standing on one leg on the back of a trotting donkey, the other leg extended out to the side. Jesus’s white robe flaps behind him, his arms are spread out to catch the wind, a look of pure glee on his face. The donkey is hee-hawing in glee. Citizens of Jerusalem look on in awe and glee.
The prompt evolved over subsequent attempts. My first attempt was with Midjourney. (At this point we hadn’t added the “entering Jerusalem” phrase.) Midjourney didn’t get “standing on one leg”, and it had problems with the animal’s legs.
Next I tried ideogram.ai. It also did not understand “standing on one leg”. At least it gave the animal all its feet.
Then I tried Gemini.google.com. The prompt now had “entering Jerusalem” and Gemini did a nice job with the scenery. It also understand the standing on one leg phrase, but it got a little excited and… behold, three-legged Jesus:
I went to Leonardo.ai. At this point the prompt had “children run behind” but Leonardo wouldn’t accept a prompt that had the words “bareback” and “children” in it. So I changed that to “citizens of Jerusalem”. Leonardo also didn’t understand bareback riding, and gave a very pedestrian image, with Jesus almost falling off the rump of the donkey.
So I went to chatgpt.com, and it got it right. While it took the longest time of any of the AIs to generate the image, 3-5 minutes, it got every detail right. It even added palm fronds on its own, so it clearly knew the references.
So that’s your pre-easter tour of the current state of AI image generators.
Codgerville 