Feb 11, 2023

The Ministry of Silly Works

Futuristic band.
 I've been experimenting with AI-generated art and discovering some of the limitations of these art-generating computer algorithms. The image too the right was generated by WOMBO Dream. The text prompt was "Futuristic band, girl band, Bangles style, four members, playing instruments, on stage".

I've previously spent many years using Daz Studio software to render images of human figures. It is easy for me to imagine quickly generating a desired pose for a character in a science fiction story by using Daz, then feeding that preliminary image into image-generating software such as WOMBO Dream.

Parthney and his band. image source
The image to the left shows an illustration from my science fiction story Exode, a depiction of a band that I made ten years ago using Daz Studio. When I made that image, I was particularly concerned with having futuristic instruments.

To go along with my Parthney band reference image, I used this text prompt: "Futuristic band, four members, one man three women, playing instruments, on stage, futuristic instruments".

Band One.
Band Two.
There were some interesting AI-generated bands. The image shown to the right illustrates one of the "tricks" that Mr. Wombo uses while creating futuristic instruments: the two-person guitar. That image also illustrates the fact that even when I request that there be a man in the band, Mr. Wombo usually refuses to include a man.

The image to the left was unusual in that Mr. Wombo allowed the Parthney to turn his back on the audience. 

Band Three.
Band Four.

You can also see that these images were generated with a background stage prop above the band members. Some of those were much more interesting than what is seen for Band One and Band Two. 

One of these more interesting background images is shown to the left. The image to the right illustrates the fact that Mr. Wombo can't count; often the AI-generated bands had three or five members.

Band 5.

Band six.
The most unique "background" image is shown to the left. Once again Mr. Wombo made a band with 5 members, and one of them is up a pole. Parthney grew up on a planet of the galactic core, Hemmal. Hemmal has a Hi Tek™ space elevator which eventually Parthney uses to depart from Hemmal. If you have a material strong enough to build a space elevator, then why not have a musician elevator like the one deployed by Band 5?

Band Seven.
Band Eight. (full sized poster)
I fed the "Band 5" image through Mr. Wombo's "Flora V2 style" and got the image that I call "Band Six". I like this version because it almost looks like a spaceship there above the band. 

One of the most strange looking musical instruments that Mr. Wombo produced is shown to the left.

The Last Band. Another odd instrument from planet Hemmal is shown to the right. Maybe those are Visi-sonors.

Now for something completely different...

Three legs backwards.
Figure 1. VFX V2 style.
One of the problems that I frequently encounter with AI-generated art is that the software can generate malformed human body parts. For example, the text prompt "John Cleese, The Ministry of Silly Walks" produced the image shown to the left. This image was made by using the WOMBO Dream "Cartoonist" style. Not only are there three legs, but the legs are not correctly assembled with the rest of the body.

Figure 2. Realistic.
The image in Figure 1 (shown to the right, above) was generated in the WOMBO "VFX V2" style. 

When I used a John Cleese reference image from the internet and the "Realistic V2" style, I got the image shown in Figure 2. Mr. Wombo seems to "think" that three feet are better than two and three thumbs are better than one.

The Palace of Love
The Star King
I previously experimented with AI-generated book covers. Figure 1 looks like it could be an illustration for a book cover. I tried this prompt: "science fiction, Jack Vance, Demon Princes, book cover, Gino D'Achille style".

The image to the right is a cover for The Palace of Love. I tried to make it include a flying aircar at the upper right, but failed.

The image to the left is a cover for The Star King. I tried to have it include a projac blaster.

The Book of Dreams

When trying to get a cover for The Book of Dreams, I included the book title in the text prompt and the AI-generated image included a book on the cover. Maybe that is Mr. Wombo's version of Alice Wroke (image to the right).

The Killing Machine
When I tried to get a cover image for The Killing Machine, I used this prompt: "science fiction, on an exoplanet, book cover, Gino D'Achille style, a cute woman riding, dressed in a futuristic jumpsuit, riding on an alien creature, the novel is titled "Thamber: The Lost World"" (see image to the left).

Hi ho heels, away!
I have to include the image to the right which illustrates how Mr. Wombo always wanted to have her riding a horse, in this case, a horse dressed in high heels.

The Face (click image to enlarge)
Here is a cover for the last of the five Demon Princes novels: "science fiction, moonlit night, book cover, Gino D'Achille style, a cute woman running, dressed for carnival, running from a spaceman, futuristic mansion in background, a novel titled "The Face"". 

This image that the software generated could conceivably be a mixture of two different scenes from the novel, one on Dar Sai and one on the planet Methel. Previously, I tried to make a cover image featuring "beauty" Dasce from Vance's novel Star King. Here, I imagine that is Jerdian on the AI-generated cover for The Face (lower right corner).

Jerdian
Jerdian again
 Speaking of Jerdian. Here are two book cover images for the first time that Jerdian notices the way that Gersen is looking at her when he sees her at Dindar House on Skansel Plaza. I asked Mr. Wombo to make a futuristic city scene, but this is not what the planet Dar Sai would look like.

I prefer the version of Jerdian in the image on the left, but I don't like the way she is carrying the documents in that image.

Figure 3. nanite vial
 Nanites. While trying to get a sensible depiction of Jerdian on the rather low-tech planet Dar Sai, Mr. Wombo kept morphing the background into some sort of cyberpunk city and the images began having what looks like nanite containment vials (see Figure 3). I'm always in the market for ways to visually depict nanotechnology.

Figure 4. another bottle
 It's Green. I really like the Hi Tek™ bottle in Figure 4, but the image to the right is made weird by the finger-like growth emerging from the palm of her hand just below the bottle.

 Figure 5. may the bird of...
I was having a good run of generating interesting looking futuristic nanite containment vials and then Mr. Wombo gave me the image in Figure 5. Mr. Wombo's "VFX v2 style" is good for nanite bottles, while Figure 5 used "Flora v2 style".

Figure 6 . future of six fingers
The image in Figure 6 also has a finger problem. I can't decide if that is a phone or maybe some other device from the future such a Reality Viewer.

Figure 7. Book Deal.
On the topic of strange things that Mr. Wombo does with hands, take a look at Figure 7. I'll call this "the underhanded book deal".

Figure 8. Thumbs down.

The "Flora v2 style" can also generate some highly pathological hands (see Figure 8, click the image to enlarge). I have no idea where the feline humanoid standing there in the background came from.

I can't resist mocking Mr. Wombo's confusion over human body parts. I specifically asked for some human hands with nothing else going on in the image. All the AI-generated hands were anatomically incorrect.

The Ugly Hand Book by Mr. Wombo.
The image to the left is from a recent book fair where Mr. Wombo's Ugly Hand Book was a big seller. 

Mr. Wombo Does Arms
The next book in the Mr. Wombo anatomy series is shown to the right. In the first edition, this book was sold as Mr. Wombo Does Arms Much Gooder, but then they shortened the title for subsequent printings. Click on the image to zoom in on the amazing nails of the two book company representatives.

Dangerous Nanites

I'm preparing to send Tyhry and Marda off to Alastor Cluster in search of nanite technology. My hope is that they can discover some useful nanites while on a relaxing vacation, but some of the nanites might be dangerous in the hands of primitive Earthlings (see the image to the left).

Figure 9. I request a change.
 The Bottom Line. Mr. Wombo is pretty good at producing derivative versions of images that are already popular in our culture. However, when I try to have Mr. Wombo help me illustrate a new idea from my science fiction stories, it can be very difficult to get what I want from the image-generating software. 

As I write this, WOMBO just switched on a new feature that allows you to use up to 450 characters to describe what you want in the scene, but this feature seems to be easily confused and usually does nothing useful for me.

Wombo style
Figure 10. My idea
I requested a depiction of a woman being teleported (see Figure 9). Then I requested the addition of a "shaft of glowing light". The image to the left is what Mr. Wombo did with my request for a blue teleportation beam; the entire scene was tinted blue.

What I had in mind is shown to the right. When I fed figure Figure 10 back to Mr. Wombo, he generated the similar image that is shown in Figure 11.

Figure 11.
Figure 12.
I then requested that the teleporting figure be given brighter clothing with sparkles. The resulting modified image is shown in Figure 12.

I find it very easy to make suggestions that make an image worse rather than better. Collaboration with Mr. Wombo is very frustrating. I suppose the software will get better with time, particularly if the software is given some actual intelligence.

Arrival on Earth
Teleported to Earth

Eventually, I got the two images that are shown here (to the left and the right). I imagine that these are interventionist agents who have just been teleported to Earth where they must complete their secret missions. I prefer the teleportation beam in the image to the left. If this is Manny the bumpha then the image to the right is a better match to her nanite-sculpted hair.

instructions for how to make adjustments to images

Space Command!
Cowgirl Command!
Mr. Wombo is better at depicting conventional images such as people just standing around inside a spaceship. These stern ladies wear their blasters mounted on their hips.

I asked Mr. Wombo to convert that futuristic spaceship scene (image to the left) into a scene from the old west (image to the right). I had to edit Mr. Wombo's image to insert the balloon.

Movie filming; 1942
European vacation
Here are two other AI-generated versions of these three ladies. The image to the left was made in "Jane Russell style". I'm not sure what that is that she is holding in her right hand.

The image two the right was generated by WOMBO's "Flora v2 style". I had to do some editing of the sky in order to remove some unidentified flying objects.

Exploring the big city.
night version
For the final pair of images, we are off to New Your City where a visitor from the country has been spending a day sight-seeing with a friend. 

For the image to the right, I asked Mr. Wombo to convert the original scene into a night time version. I'm not sure how Mr. Wombo decides how dress the people who appear in the images that are generated. One of the AI-generated images had an appearance like a character from Star Trek, so I went along with Mr. Wombo and imagined an new film called The Voyage Home 2.

San Francisco
what rank?
By the wonders of time travel, the image to the left shows 23rd century fashion on the streets of San Francisco.

Mr. Wombo really got into giving promotions to these Star-fleet officers (see the image to the right). Maybe her rank is Ultra-Superlative Admiral.

Next: more Sci Fi scenes illustrated with AI-generated images.

visit the Gallery of Movies, Book and Magazine Covers


No comments:

Post a Comment