AI Art College Quickies Posted by Ian Symes on 7th June 2023, 09:35 The new Beta version of Adobe Photoshop has an interesting feature called Generative Fill. You highlight a section of a canvas, and it uses AI to figure out what should be there, based on the rest of the image. Immediately, people started using it to extend famous artworks, album covers and the like. My immediate thought – what happens when you put Red Dwarf images through it? This image, which just happened to be on my desktop, was the first one I tried to extend. The results were significantly more disturbing than expected. Then, I figured out how to actually use the tool. As we all know, the original version of Kryten was a young Black woman. Here’s a cropped version of a famous publicity picture… …and here’s how the AI imagines their legs to look. You can also get it to fill in gaps within the original shot, such as this screengrab of a headless Kryten. Err… has anybody seen my legs? Yes. Ever wondered what Holly’s body looks like? Ever wondered what the other Holly’s body looks like? Ever wondered what Cassandra’s body looks like? Ever wondered what Pree’s body looked like? At this point, we were attempting to get the AI to show us a bit of cock, but it wasn’t playing ball. It did however answer the question of what exactly Rimmer-in-Lister’s-body was so shocked to look down and see. Double polaroid sadly covered up. This is what the AI reckons Ziggy’s costume looks like, if you only show it the top bit. And apparently the Polymorph looks something like a roast chicken. Just relaxing on the beach. This obviously invites comparisons to Lister’s new body at the end of Nanarchy… …and the AI gives us a rather more realistic alternative to that too. Time for an experiment. How does the AI “fix” each half of Miranda? This is pretty clever. Adding in the surface of the water and creating a very decent reflection. This, on the other hand, is horrifying. We fed it the opening shot of The End, in the hope of recreating the abandoned new effects shot revealed in the Re-Dwarf documentary. The AI’s imagination does not match that of late 90s Doug and Ed. Exclusive – revealed for the first time, this is what Lister’s dad looks like. And this is the bottom half of his gran. Just to complete the set. I call this series “You CAN Move Outside The Edge Of The Photograph”. Now to give it a real challenge. Fill in the gaps here… O… K. And finally, the classic. Uncrop. These conventions get weirder every year.
Disturbing stuff. Although the bodies for Holly/Cassandra/Pree are pretty good. Queeg and Gordon next!
“He had the whole building transported brick by brick to the surface of Mars, just to get away from the neighbours”.
Crikey, it was good of Hattie to make such an effort. Those aftershow parties must have been quite something.
Stupid sexy Cassandra. This AI image expansion tool is clearly very powerful in the way it made all these accurate extrapolations (such as the huge crowd at the Aigburth Arms to see Smeg and the Heads). Obviously you should use it to convert the whole BBC era to 16:9, for Smega-Drive gif purposes.
I love the attention to detail in that it’s given the crowd at the Smeg and the Heads gig a touch of the crowd at McIntyre’s funeral from Remastered look.
Maybe Rimmer didn’t have to be so jealous of his siblings, Frank’s getting married in an absolute dump!
I’ve only ever heard the ‘AI can’t do hands’ trope, today is the first time I’ve seen it. The Holly bodies are a rather disturbing demonstration O.o
Thanks for the nightmare fuel. I’ve only ever heard the ‘AI can’t do hands’ trope, today is the first time I’ve seen it. The Holly bodies are a rather disturbing demonstration O.o I like how it seems to have given up on Pree’s and just hidden them behind a big black splodge.
This week on G&T: Cappsy starts a thread worthy of being its own article, and Ian Symes publishes an article that reads as a glorified thread post. Good stuff, though.
Pitch: new Smegle mode where you have to guess the episode based an the AI extrapolated image, but the original frame is obscured until you guess correctly.