Midjourney (version 4) proved to vastly enhance a lot of my bad linocuts with AI:
LITERAL ENHANCEMENT
I uploaded an image of my bad print with a more literal descriptions of the actual scene:
Midjourney prompt: "black and white linocut print of a woman sitting on the floor"
Midjourney prompt: "linocut print of a woman sitting down"
Midjourney prompt: "linocut print of a woman sitting down"
Midjourney prompt: "linocut print of a woman sitting on a couch"
LESS LITERAL
I fudged a bit by entering prompts of what the bad linocut suggested, even though I didn't draw that exact image to begin with:
.png)
.png)
.png)
Midjourney prompt: "linocut of a woman in front of a traffic intersection"
Midjourney prompt: "linocut of a man standing in front of a corn field"
.png)
MIXES
A Mix of the previous two AI generated linocuts, no prompts:
Mix of two images
To stir things up I first generated a fake AI print in the style of the Mexican Master printmaker Leopoldo Mendez with only a prompt:
Midjourney prompt: "linocut print in the style of Leopoldo Mendez"
.png)
Then I mixed the best generated result with one of my bad linocuts (no additional prompt) and got the seated man below:
Mix of two images
.png)
FANATASY
disconnected prompts
I added a prompt which had nothing to do with the image I uploaded
Midjourney prompt: "black and white linocut print of a javelina"
Midjourney prompt: "linocut print of a rhinoceros"
Midjourney prompt: " black and white linocut print of a chupacabra"
Midjourney prompt: "black and white linocut print of a desert tortoise"
Midjourney prompt: "black and white linocut print of a cricket"
What remains to be done is to laser engrave the best generated images into a 6x6 inch square of linoleum at Quelab (see settings) and make prints for the BUCKET EXHIBITIONS, like I did earlier this year.
My uploaded prints seem to have a big influence, but the results generated are very different from my prints. How much of the fake print is me, and how much of it is the computer? I think uploading my prints is a very different thing from just generating entirely from text.
LASER ENGRAVING
It took about an hour and fifteen minutes to laser engrave the first image of this blog post onto a 6x6 inch linoleum square at Quelab on December 13th.
The next day (December 13th) we printed the AI linocut at Henry Morales' studio, with sepia and yellow inks:
Sepia and yellow inks
No comments:
Post a Comment
Note: only a member of this blog may post a comment.