Cloning My Dumb Voice
A very weird edition of Webworm and I'd love your feedback.
Hi,
New Zealand auction site TradeMe is still giving conflicting reasons for why it removed the gorgeous painting of Prime Minister Chris Luxon. It took a few days, but Webworm’s story spread to RNZ and the Herald this week. I’ll keep you updated.
Today is going to be a very self-involved Webworm — and that’s saying a lot because there have been some very self involved Webworms.
Face blindness. Adult ADHD. Squirrel bites. My terrible memory.
That sort of thing.
This one is self involved because it’s about my voice. I will try and keep it brief.
I have been toying with the idea of doing an audio version of my written Webworm stories. I do this sometimes already — like reading out Josh’s piece on climate change, or my piece on Gary — but I have discovered that for me, the joy of Webworm is me perfecting something I have written and then immediately hitting “send”.
A burst of immediacy and adrenaline. And excitement at reading your feedback.
Recording a voiceover to a written piece means another layer that my slightly spicy brain finds very stressful and frustrating. If I mess it up, it means I have to edit the audio, then I find pops and things I don’t like and I end up editing more, then suddenly hours have gone by and I haven’t sent it out. Plus if I make an error, it’s easy to amend in text — less easy in audio.
(I’d note this all differs to episodes specifically as podcasts, like Spaceman Barry. That’s a different thing).
So I parked the idea.
Then a few weeks ago I was on a long drive, and I noticed the New York Times had an option where an AI voice could read articles to you. I did it, and I mostly liked it. It meant I could absorb information while I drove. It felt like a bonus feature. It was a safe way of “reading” the NYT while I concentrated on staying alive behind the wheel. American freeways still freak me out.
And so — despite disliking many aspects of AI — I started thinking about audio versions of written Webworm pieces again. Even just from an accessibility point of view.
I then found myself talking to my friends from a band I love, Kumo99. They’re huge audio nerds and they told me about voice cloning — where I could feed a bunch of my voice into a computer and it would spit out a clone.
And so for the last few weeks, I’ve been fiddling around with it. And annoying my friends with it:
Today, I’ve compiled a bunch of what I’ve been doing — and I wanted to feel you out about the results.
It may be a straight NO — you hate the idea of any kind of “computer voice” reading my stuff. Or anyone’s stuff!
But if you are open to it and would find it a useful addition to Webworm, I want to know a few things.
First, listen to the wee podcast that is attached to this newsletter:
In it, I run you through some options I’ve been thinking about — including two different clones of my voice (trained on different material).
I also run you through three generic AI readers. All of those options read the following script:
This is David, albeit an AI version of my voice. And this is a story from my friend Elle Hunt. The other day I was chatting to a friend and made a throwaway comment to dial-up internet. Her face was blank. I persevered, imitating the screechy sound of the modem, the fractious daily back-and-forth you’d have with your parents about using the phone. I’d never thought much of our age gap, but that ritual, it turned out, was before her time. I didn’t dare tell her that I used to check my bank balance by going into a phone box, and write down directions to my destination before I left the house.
After listening, I’d like to ask you a few things. Consider this a reverse to this week’s AMA!
- Are you open to an AI/computer voice? Yes or no.
- If you are open to the idea, do you prefer a clone of my voice, or a generic reader?
- If you prefer a clone — what option: Clone 1 trained on scripts, or clone 2 trained on freewheeling studio sessions?
- If you prefer a generic reader — what one do you like? Andrew, Matilda, or Brian?
- Or do you hate all of this and banish me to the lower depths of hell.
This is an intellectual exercise as much as a practical one.
I go back and forth on the whole thing. I woke up one morning and was like, “That clone voice sounds like me. I can’t tell the difference. I trained it on me, it’s owned by me — so obviously that’s a good option!”
Earlier today I was like, “No one wants a clone of me that’s just weird. Is it unsettling uncanny valley stuff maybe? Does it feel like I am tricking people somehow, and I should just have some generic AI voice reading it instead?”
Or — “All AI sucks, I hate it!”
Or — “I found a voice reading the NT Times useful, so why not do it for Webworm as an added option? Just so it’s there. It can be ignored if people hate it?!”
To be clear, if this clone thing was ever used by me, I would clearly tell you. I think that’s incredibly important. This is more just a case of “what is the best reader — sorry, listener — experience?”
See you in the comments!
This is weird!
David.
PS: All the free tickets to me + Tickled in Sydney and Melbourne went very quickly. The cinema has some for sale if you missed out.
