Friday 19 August 2022

DALL-E 2 and other digital manipulations

  (my drawing in AR)


On August 18, 2022 I got access to use DALL-E 2.  Twenty four hours later these are my best creations, based on one of my drawings:




Three layers of AI








PROCESS

While DALL-E 2 is a "text-to-image" AI generator, it will allow one to upload an image.  I uploaded a square drawing that I did on July 28, 2022.




I erased the face of my drawing and put in the text "face of a beautiful woman."  DALL-E 2 "inpainted" a face (in the erased part) and gave me three results..  I choose the first result, the image below.  These steps are not show in pictures in this blog:


DALL-E 2 gave me this image


I asked for variations on the image above, and again DALL-E 2 gave me three choices.  I chose to develop the first altered image:

I choose to develop the first altered image


The first altered image
of the above


I ran the altered image through Deep Dream Generator, which applied a new style and color to the result (I then further edited it in Photoshop Elements):



Then once again I uploaded the latest revision to DALL-E 2, and "erased" most of the figure:
 
I uploaded and erased the figure
in DALL-E 2


Again, DALL-E 2 "inpainted" a new figure, keying on the text I typed in "a beautiful woman in a bikini."  DALL-E 2 gave me three results:

"a beautiful woman in a bikini"

The final images do not look much like my drawing.  However note that this last step conserved a lot of the style of my previous input image, rather than just superimpose a photograph of a random woman in a bikini.




Earlier DALL-E 2 attempt


Earlier Holly Grimm gave me one of the DALL-E 2 creations she generated with the prompt -- "black and white print of an artist working in the style of Carlos Barberena,"  referring to a great printmaker in Chicago. I engraved one of the results on a linoleum square using the laser cutter at Quelab, and then printed the image on a Tortilla Press.  Then I framed the 8x8 inch relief print and it came out quite nicely:






**Midjourney**

Note: I tried the same prompt in Midjourney  -- "black and white print of an artist working in the style of Carlos Barberena."  Midjourney is the competing AI art program to DALL-E 2 right now:






So I used the same prompt in DALL-E 2 -- "black and white print of an artist working in the style of Carlos Barberena" -- and got three entirely different variations.  



Three new and different variations of
"black and white print of an artist working in the style of Carlos Barberena"


The first new variation was the best, but as a drawing, not so much as a print:

First variation



Second variation


Third variation


I ran the first variation through Deep Dream Generator, to beef it up a bit:




Editing the variation


I used Deep Dream Generator to make an even beefier version of the DALL-E 2 line drawing:



HOLOGRAMS

Then used the two "beefier" DALL-E 2 images to make holograms viewed on the Portrait display by Looking Glass Factory.

First I had to make "depth maps" of the images, and place them to the right of the image in one JPG. I used DPT Large to generate the depth map:






The AI holograms turned out nicely:


AI hologram
(behind the AR image)





Web AR

We also experimented with Web AR at Quelab last Wednesday (August 17th) -- and it worked!  One could see my 3D drawing in Augmented Reality (AR) WITHOUT downloading an app.  Adric could see it on his Pixel 5a phone, and Robert could see it on his iPhone 12 (though the figure lost volume on the iPhone).  The older phones will not work, so the key is to use a newer phone with AR hardware:


My AR drawing
floating in real space


Brian posted both the code and my 3D drawing on his server.  You can experience the AR by going to the links below on your newer smart phone:


Apparently one has to point his smart phone at a clean flat horizontal plane and my AR figure will pop up. It might be very small, so one can enlarge it by un-pinching their fingers.  I believe that the viewer can also place the AR figure anywhere he wants.

This Web AR (augmented reality) does not allow for "targets."  That is, one cannot point their smart phone at a print, to have that print generate an AR image.  We did do this however using prints as "targets" in the Desert Triangle Print Carpeta:





THOUGHTS


I always want to push art further into the 21st Century.  I would love to combine AI, AR and holograms into one experience.  I would love for one to scan our framed 8x8 inch prints with their smartphone and enjoy an additional AR experience.  Perhaps there is a way to let the hologram display know which print is being viewed with AR, so that same print could be displayed in three dimensions on the Looking Glass Factory hologram display.





OTHER AI

There are other new AI art programs coming out now:


Plus there are some other interesting digital art tools surfacing as well:






Afterword
Other Examples

 TAPIR

Created in DALL-E 2 and altered in Deep Dream Generator.  Further edited in Photoshop Elements:




Hologram ready --
image to be loaded into the 


WebAR

The tapir was removed online from the image above with RemoveBG, then imported into Monster Mash to be outlined and create a 3D file.  The resulting GLB file -- tapir.glb -- can be downloaded  online.



AI Tapir by krrrl on Sketchfab




Flying Tapir from AR (Augmented Reality)
during Quelab Open Hack
on August 24, 2022



For creating WebAR, the zip file -- Tapir.zip -- can be downloaded from THIS FOLDER online.  If all the contents of this zip file were loaded to a server, I believe that one could see the Augmented Reality tapir in 3D just by going to the website.  No app necessary to download.  However the viewer must have a newer smartphone, like a Pixel 5a or latest iPhone.



HACKERSPACE BOOGIE

Robert Atkinson printed out some of his AI creations he made in Midjourney:




No comments:

Post a Comment

Note: only a member of this blog may post a comment.