AI Art Adventures: Fiddler on the Ref (Ref’ing on Classic Art)

I was going to call this the Art Forgery Edition, but realized that wouldn’t go down too well.

My post on the 17th century artist Diego Velázquez made me think. What if I took one of those imaginary lost paintings and used Midjourney to create it, using –sref and –cref parameters? As a plus, he’s long dead, so no one can accuse me of ripping off someone’s artistic creations.

So let’s create Two Dwarves Posing in Armor Before a Fountain. 

Here’s the first image using just the prompt “two dwarves posing in armor before a fountain, painting by Diego Velázquez.” Not bad, the AI is clearly drawing on the many Velázquez paintings floating around, as well as, I’d guess, other 17th century ones. The armor is clearly Spanish Baroque style (think Conquistadors) and the men are clearly dwarves, with adult heads but small hands and feet. They are clearly posing. But let’s select a more specific painting for reference and insert its URL right at the beginning of the prompt. Lucky for us Velázquez supplied us with several dwarf paintings so I’ll pick one.

Portrait of Sebastián de Morra, Diego Velázquez, 1644

This fella was employed as a jester in the Royal Spanish Court. His look says it was a rather ambiguous position and he wasn’t treated that well. Velázquez himself considered dwarfs fellow humans and not animals or pets, but most of the royals weren’t that enlightened.

Let’s see what happens with a URL reference: “https://s.mj.run/9LfsE0BHxqs two dwarves posing in armor before a fountain, painting by Diego Velázquez”.

Right away we have a difference. The brushstrokes are more notable and feel more authentic, but the dwarves have also aquired the pointy ears Midjourney wants to put on every humanoid related to fantasy art. The armor had been lessened to cuirasses and some bracers. Red appears in the picture from the jacket of the reference dwarf.  Their facial expressions appear more lively, but that may be me.

The same prompt with Velásquez’s original now used as an –sref (“two dwarves posing in armor before a fountain, painting by Diego Velázquez –sref https://s.mj.run/9LfsE0BHxqs”) Note one set of pointy ears is still there and the rim of the fountain is… weird. There’s no spray and the ledge wobbles slightly upward. Also note the dwarves are looking more alike, as if they’re becoming twins. I do like the deep colors of the clothing, and the cuirasses still seem authentic (though I’m no expert.)

With a –cref, “two dwarves posing in armor before a fountain, painting by Diego Velázquez –cref https://s.mj.run/9LfsE0BHxqs”. Very nice and authentic looking. You can see how the original dwarf’s face has been replicated, used as a model. Both have large, square heads, small feet and hands with no extra fingers. While their ears are still slighty pointy it is within the human norm. There are subtle signs of pride on their faces as they pose, and the dwarf on the right seems to have a pistol of some sort. In back of them is a fountain, but there’s no water. Oh well, you can’t have everything. (Did I just make a pun?)

Which one gives the best result, that’s up to the user. I can’t help but like the first one as the armor is more extensive and there’s a gushing fountain, but it doesn’t scream “VELAZQUEZ!” like the –cref one does.

What happens if I use the last prompt, but replace dwarves with dogs?

I love the way they are staring at each other as if thinking what the hell? in mutual puzzlement.

The character at left COULD be a dog with some styling, his doggie haunches hidden by the voluminous sleeves, or a doggified dwarf. There’s no way of telling.