Artificial Intelligence is a bit of a topic these days. Questions like, what is that? How does that work? Am I out of a job now? Is this from Satan? I can’t answer any of the above, but I have recently been paying and playing with AI for my own purposes, so if you don’t really understand any of it, I am here to help, not with any factual details, but as someone who has been at least an AI passenger to Anna, my patient wife, who was herself helping me through it all and doing the buttons.
If this is already sounding complicated, then please, I assure you, we are not even getting started. So where best to start from?
I’d liked the thought of creating audio books of my novels since I’d written the first one. There is something tangibly close about being read a story and I would often posit the idea of finding someone to read mine into a mic.
‘Reading out loud, that’s basically all that it is. Simples,’ I would say to Anna, and I even had someone in mind to do the reading. Of course I did. I am not going to say his name, or even his more familiar name that me and my extended family know him by, but he had a significant background in both English Literature and Truck Driving. More importantly, he also has one of those warm, welsh voices which I was certain would fit lovely with The Meifod Claw.
‘I think you should try recording yourself first,’ Anna would say back, reiterating that it might not be as easy as I was making it out. So I did record myself reading in my garden and it was terrible. Awful. But my trucker man had one of those Masterchef’s in English Literature, so when I did ask him, I did so confidently, even as he was shrinking away from the idea. Like me, he didn’t favour the sound of his own voice, and also like me, he just sometimes wouldn’t do what he was being told to.
Stalemate. And that was where I left the audio book idea. It would still linger around in the Good Idea cabinet of my brain, but its’ time wasn’t yet… I assumed it was waiting for me to have the available cash to pay someone, and thus get my demands that way, but it turned out that the idea was waiting for technology to begin getting half decent. Maybe.
Those elements began to come together about three months ago when I was going through a round of editing on the first part of this fantasy series which I had started. One story written through ten small books. It is a format that I have been enjoying, and quite enough work in and of itself, but lo, some penny dropped and I did a thought that approximated, AI can read things that you feed to it… what if I fed it one of my books? How would it read that?
Anna and I had already been using a very rudimentary computer reader for a few years as an editing tool; it does not read the text with any feeling, but that does counter-intuitively help with finding errors. But modern sophisticated readers promise to go better than that… they even say that they are intelligent. Intelligent enough to understand and uphold the tone of reading a whole novel? I was going to give it a shot, but also be realistic; rather than feeding it a whole novel, I would perhaps feed it the first part from the series I had recently started writing. Everyone loves a fantasy fiction read aloud to them, don’t they? That’s not realistic, but it is what I hung the idea on once I had decided to start another round of audio book discussions with Anna. I couldn’t really rely on discussing the actual details of AI readers, because that was the sort of enthusiasm that I was trying to drum up in her to find out. The weights and measures of computer variables are utterly, or else entirely, above me; they have no weight or interaction as life does otherwise, nor else some tangible experience of something spiritual. A terabyte of data is something that I have no context around. I know that it is a thing, but…
So I was really going to need Anna to join the circus with me on this one. And it must have been a good idea, because she span around, into her producer’s leotard and went off looking for that perfect balance between the two things that we look for in this type of situation; the very good and the possibly cheaper. A day later I was presented with a one-off monthly subscription to Natural Reader and their low tier selection of taster-readers. I felt like I had been propelled into a dating app, and after some quick spooling through the available voices, I had found my first AI reader date. His name was Onyx (stop laughing) and I was genuinely interested in meeting him in an early edit of Border Vaudeville. Onyx sounded broody and the text was gothic…
And for the first few minutes of listening to Onyx read, I was genuinely impressed with how he (it? thing?) was doing. Like a good opening song from an album, he was hitting the correct tone and getting me hopeful for what would follow.
Then the first inevitable snag… Onyx doesn’t know one character from another character. My lead pair for Border Vaudeville are Clyde Barrbough and Rubus Rhume, but Onyx does not know that. You or I might have a sense of delivery for each character’s speech but Onyx doesn’t. Onyx just reads, each word as distant as it was understood. It occasionally meant that if his tone was there, the melody of delivery was all over the place. It didn’t jar obscenely, but when I encountered it it was like when Anna would put new curb marks on various old alloys. Not world ending but it drew attention to itself and really shouldn’t have been there.
But by a good margin I was impressed with Onyx. There were entire sequences of the text which were read out as if a person were in the room with me. I could see the broad potentials of it, and that had been before Anna had started playing with settings and crunching changes down to the fractions of seconds on the equaliser.
Because here’s the thing when you begin to go deep with AI… you gotta hold its hand. Not out of romance, my goodness no, but rather necessity. You are going to have to smooth over the bumps.
Editing is a great process, if boring to describe. And if anyone even wants to know about your writing in the first place, they certainly have no ear for the finer points of the final process. Editing isn’t mechanical, but it has a melody that sounds like you are describing an HR department.
Anna and I spent an entire day in the mechanics of that department some Saturdays back, herself armed with the buttons and a quick witted know-how, and me with the interjections and opinion because I had written the thing in the first place. Both the barrels loaded.
And somehow it does work. In the space of a clear day we choose a final voice reader called Vincent and went about managing changes to his enunciation and pacing. Onyx had been plenty serviceable, but now that we had paid for the full suite, we had choices of voices coming out of our ears; well into them, and mostly they were not what we were listening for. Vincent though, he sounded like the Universal Grandfather, so we copy and pasted the text to him and editorialised the result before I hassled my artist elder brother for a picture appropriate for a YouTube thumbnail.
Simples, and what it has all left me feeling is that you can never get away from the need to smooth out and edit things, regardless of the sellers promise, and it is in fact that process which yields something like a decent result; the best you could have reasonably shot for perhaps, if such praise is yours to say. Basically I am pretty happy, but I think this is because between myself and Anna, we have sanded the grain of this thing until the paper ran thin and our fingers began to get hot. Well, her fingers as she was doing the buttons.
So, I do not know if AI is evil or whatever else because I do not know the clearer picture of its applications. It can basically do reading, if that is a measure. At the moment I can say that AI is like a village simple man; he might not have the smarts, but if you get him down the farm, he sure can lift stuff and shovel hay, and no-one doesn’t see a use for that.
God bless it I say. Plus a thank you to Anna. And God bless her too, absolutely.