PROCESS
1) VQGAN+CLIP -- I started last year (2021) by generating fake de Chirico architecture with a VQGAN+CLIP program in Colab:
2) DALL-E 2 -- Then I expanded that scene, twice, with DALL-E 2:
DALL-E 2 prompt: "Unusual architecture in the style of Giorgio de Chirico"
I further "outpainted" this scene in DALL-E 2
I previously made two other videos featuring de Chirico Architecture:
3) Midjourney -- The rest of the images I generated in Midjourney (Version 4), almost always using the prompt: "in the style of Giorgio de Chirico."
MIDJOURNEY
a) I uploaded the second image above with the prompt: "in the style of Giorgio de Chirico," to make the first batch of images:
One of the resulting images
generated in Midjourney
b) After converting one of my drawings into architecture using VQGAN, I uploaded the image result with the prompt: "in the style of Giorgio de Chirico" :
From the first generation batch
From the second generation batch
c) I added my figure drawing to another building I generated in AI (based on a different drawing), but Midjourney would not create any variations, claiming that my figure was NSFW. So I changed the color in Palette FM, and then Midjourney allowed me to upload the image and continue. I also entered the prompt: "in the style of Giorgio de Chirico":
From the first generation batch
From the second generation batch
From the third generation batch
From the first generation batch
From the third generation batch
e) Yet another building generated in VQGAN based on one of my figure drawings, with the prompt: "in the style of Giorgio de Chirico":
From the first generation batch
f) In order to avoid the NSFW rejection by Midjourney, I added a generic Running Man figure, to a building I generated in VQGAN (from my drawing):, with the prompt: "in the style of Giorgio de Chirico":
August 31, 2021 -- Figure drawings into Architecture
From the first generation batch
From the second generation batch
g) I added the Running Man figure to the first generation image (using Photoshop Elements), with the prompt: "in the style of Giorgio de Chirico"
Running man added
to the first generation picture
From the first generation batch
From the second generation batch
h) Again, I generated another building from one of my figure drawings in VQGAN and uploaded it with the prompt: "in the style of Giorgio de Chirico":
From the first generation batch
i) Running Man added to a first generation image, with theprompt: "in the style of Giorgio de Chirico":
From the first generation batch
From the second generation batch
j) I took a picture of the coffee shop and also uploaded a fake de Chirico image, and let Midjourney blend the two:
From the first generation batch
k) I took a picture of the Quelab hackerspace at sunset, uploaded it to Midjourney with the prompt: "in the style of Giorgio de Chirico":
From the first generation batch
From the first generation batch
l) I took a photo of the Sandia Mountains at sunset in Albuquerque, uploaded it to Midjourney and added the prompt: "in the style of Giorgio de Chirico":
From the second generation batch
From the second generation batch
From first generation batch
PALETTE FM
I colorized the image above in Palette FM, which delivers many variations:
No comments:
Post a Comment
Note: only a member of this blog may post a comment.