More (free) AI creativity tools

 

The above may not be great comic-book art, but it's passable enough, and (with a couple of caveats I'll come to later) I created it with a single mouse-click on a website where I didn't even have to log in, let alone pay any money. And until I press "Publish" on this post, no one but me has ever seen it before. Even after a year of playing around with AI software, this still strikes me as incredibly cool (and even slightly frightening, if you stop to think about all the implications).

This particular page came from the Comic Factory website, one of two new AI discoveries I want to talk about in this post. First, however, I thought I'd summarize a few things I've written elsewhere about the "creative" use of AI. This is becoming a major hobby of mine, and I suspect it will feature increasingly often on this blog. Don't worry, though - all the applications I'm going to talk about have at least a tenuously fortean connection.

To start with, there's the article I wrote for Fortean Times last year called "AI, Art and Forteana" (FT 433, July 2023). This included a couple of imaginative fabrications from Bing's AI chatbot: first, an account of the Roswell Incident written in the unmistakable style of Charles Fort (who actually died 15 years before Roswell), and then a (hopefully entirely fictitious) conspiracy theory about Fortean Times itself.

The magazine article also contains two pieces of artwork courtesy of Bing's Image Creator. One takes the form of a two-page comic (with each of the panels having been generated separately, and then put together and captioned by myself), which also turned up in a guest post I did for Kid Robson's "Crivens" blog called Nostalgia Meets Modern Technology.

The other piece of art included in my FT article was "an engraving in the distinctive style of William Hogarth, 'proving' that he witnessed a UFO hovering over the streets of London in the 1730s". The same image also appears in a long post I did on my professional blog, 6 experiments in creative AI. As the title suggests, the "Hogarth" image was just one of several experiments discussed in that post - to read about the others, just click on the link.

Although I've occasionally found Bing's chatbot useful as an "ideas generator", the most impressive use I've found for it is in writing song lyrics, which it's surprisingly good at. I've put a couple of examples on YouTube:

• Zen Matrix - a mystical/hippie song that I think is really great (so you probably won't click on it)
• Demonic Tarot - my attempt at a "blackened death metal" song in the style of Behemoth, which really isn't very good (so you probably will click on it)

As well as AI-generated lyrics, both those videos use AI-created artwork - as do most of the recent videos on my YouTube channel. And as for this blog - you can see how often I've used Bing's Image Creator by looking at "AI art" in the tag cloud in the right-hand sidebar.

Besides being free, Bing Image Creator is also impressively high quality. But it has disadvantages, too - the biggest being that it has no memory from one image to the next. This limits its usefulness in creating comics - hence my decision to check out the AI Comic Factory. To ensure full consistency of characters and locations you need to set up a paid account, but there's a free "playground" where you can try it out without doing that.

The prompt box says "story", which seems to invite you to type quite a lot, but I just settled for a two-word prompt, "ancient aliens". The resulting artwork is noticeably less sophisticated than Bing, but it does have the advantage of being laid out like a comic - and, despite being the free version, having some vague continuity from one panel to the next.

There were just a couple of things I didn't like: the page had a broad 4:5 aspect ratio, rather than the 2:3 of a normal comic, and one panel was in a jarringly different style from the others. So, to produce the version you see at the top of this post, I deleted the dodgy panel and rearranged the others into the correct aspect ratio. But contrary to what you might think, I'm not responsible for the clumsily cropped right-hand edge of the lower panels. That was the computer's fault, and it looks like a bug to me (as opposed to the "unreadable" text captions, which I don't think are a bug, but just placeholders to add your own words).

While I won't be signing up for a Comic Factory account, I still think it's great that they let you play around with the free version (which it's best to do first thing in the morning UK time, before the server slows down as more people start to use it).

A point that was raised in the comment thread to my aforementioned Crivens post was the role of AI in creating last year's "new" Beatles song, "Now and Then". Specifically, AI was used to separate John Lennon's voice from the instrumental accompaniment on a demo tape. This brings me to the second thing I wanted to talk about in this post - because, as of a few weeks ago, exactly the same functionality is now available to anyone, free of charge, as part of a new toolkit called OpenVINO for Audacity.

I've tried a few experiments with this already, including Ozzy Osbourne singing "Paranoid" with the band muted and replaced by a classical string quartet, and a groovy Sgt Peppery remix of "A Hard Day's Night". Unfortunately, copyright laws mean the wider world is never going to hear them! Instead, you'll have to settle for my electronic remix of the Queen of the Night from Mozart's Magic Flute - the fortean credentials of which I discussed in a 2013 post on Fortean Opera (I scarcely need to add that the image in the video is c/o Bing Image Creator):

Incidentally, I'm always puzzled as to why some of my YouTube videos do much better than others in terms of views. For example, the totally brilliant Very Model of a Modern UFOlogist has only collected 20 views in the 2 weeks since I posted it, and "Zen Matrix", which I also like, has had a mere 56 views in 11 months. On the other hand, "The Doctor Who Theme in the Style of Handel", which I did at the suggestion of regular follower Colin Jones a month ago, has had a very respectable 460 views in that time - and now my version of Mozart's Queen of the Night is already well past 400 views only four days after posting it. Oh well, I guess it's just that "Doctor Who" and "Mozart" are more popular search terms than "Zen" and "UFOlogist".

UFO song #2

 

Image by Bing AI

The album pictured above doesn't exist yet (and probably never will), but at least I've now got a second song for it. It's basically an electronic arrangement of Gilbert & Sullivan's "I Am the Very Model of a Modern Major General" from The Pirates of Penzance, with reworded lyrics covering all the clichés of modern ufology that I could think of. The synthesized vocals aren't always clear, so you need to see the words on the screen to follow them, but apart from that I'm really quite proud of this one. So do please have a listen to it before reading any further!

A typical song has 8 syllables per line, with pairs of lines rhyming on the last syllable. But this one has 16 syllables per line (which isn't a problem) and rhymes on the last three syllables - which is a problem, at least for me. In the original by W. S. Gilbert, he makes it look easy, with rhymes like "lot o' news ...  hypotenuse" and "Sir Caradoc's ... paradox". In my version, I was quite pleased with "Zeta Reticuli ... particu'ly" and "there was a clash ... saucer crash", but a lot of the others are distinctly strained (such as "correct approach ... rectal probe" and "disastrous course ... astronauts"). Then again, I think there's a sense in which these almost-but-not-quite rhymes add to the humour!

As I often do with these musical experiments, I beta-tested it on five friends - and this was the first one that's ever come back with five "yeses". One person (who liked it enough to share it on Twitter/X, despite describing an earlier song of mine as "tuneless wibbling") raised the interesting question of what a "ufologist" should be called, now that the US government has deprecated UFO in favour of UAP. I suspect the answer is that UAP will very soon (if not already) come to be associated with "debunkers" who seek to explain sightings in non-extraterrestrial terms, leaving UFO for the true believers - who will thus always be ufologists.

Also on the subject of feedback, I can't resist including this nice comment from my cousin - I never knew such specialized emojis existed!

The singing synthesizer I use is a Japanese one called Sinsy, which is actually a deep neural network (i.e. AI) that gets better the more it's exposed to training data. It's been updated a couple of times since I started using it in 2020, and has improved noticeably in that time. For example, the synthesized voice doesn't sound nearly as good in my previous UFO-themed song, which I did in 2022 for my book The Science of Music.

One of the topics in that book is "algorithmic composition" -  specifically the use of Markov chains. If that's something you've never heard of, it simply means analysing a piece of music to work out its "transition matrix" - the probability that any given note will be followed by any other particular note - and then getting a computer to generate a new piece of music using the same transition probabilities.

Since the editors wanted a focus on well-known music, the transition matrix I used came from Ed Sheeran's "Shape of You". This was something of an in-joke, since he was busy extricating himself from a plagiarism accusation at the time, so I thought he wouldn't mind if I borrowed his transition probabilities (the fact that his song only uses five different notes made the maths easier, too). For the lyrics, I dug out some UFO-themed limericks I wrote ages ago, and called the result "Mark of (Markov - get it?) the UFO". Apart from the rather strange synthesized vocals, I think it sounds quite good. Here it is:

Is Anyone Out There?

Is Anyone Out There?

It seems a fair enough question, as it's 7 years since I last updated this blog, so I'll be interested to see (from the viewing stats and any comments that may come in) if anyone actually sees this post. I've been far from idle in those years - I've had 5 more books published by Springer and another 5 by Icon Books, as well as writing over 70 magazine articles - mainly for How It Works (where I've contributed to almost every issue for more than four years now) but also Fortean Times, All About Space and even (my most "up-market" moment) BBC Science Focus magazine. I never actually stopped blogging, but only in the form of relatively brief announcements on my writing blog.

I've recently become a great fan of Bing's AI tools. For example, just now I typed in "A humorous UFO or alien themed cartoon that could be captioned Is Anyone Out There?" - and was presented with the picture at the top of this post. I didn't ask it to include a speech bubble, but it did anyway (its spelling was a bit off, but all the right letters were there so I just had to rearrange them).

Although I've drifted away somewhat from Forteana in the last few years, I still contribute to Fortean Times occasionally - including an article about AI and creativity in July this year. That was also the subject of the longest post I've done on my other blog, as well as a guest post I did for Kid Robson's comics blog. I was very grateful to Kid for the chance to do this (together with another guest post showing off some of my 1970s Marvel comics) as it started me thinking about the possibility of resurrecting this blog.

When I came to dust it off today, I was embarrassed to see no fewer than 18 comments (spread over the last 7 years) still awaiting moderation. I assumed the system would automatically notify me of these, but apparently that's not the case (I've now belatedly found where you have to tell it to do this in the settings). So apologies to all those people who had to wait years for their comments to appear!