I recently uploaded three hours of my voice and discovered the wonders of AI voices. Now, with just the push of a few buttons, I can hear a version of my voice generated by AI. It sort of sounds like me, but with a slight British accent.
My Real Voice
My AI Voice
Forty-plus years ago, I voiced the productions we were filming and editing for the Australian Navy. Back then, video production was in its infancy, and my bosses, in their wisdom, sent me off for voice training. I still thank the universe for that. Since then, I have done umpteen courses in voice, presentation, and drama, and I also teach those subjects.
My last voice job was for Nestle in Myanmar, as well as a series of documentaries for Thai PBS and Discovery. That was a year ago. The work has dried up, but now that I am semi-retired, I am not too interested in going out and chasing up work. So, I have been writing. In the past three years, I have written three books. A friend commented: “You should voice them.” Ahh…I could not think of anything worse.
Audio Book – It takes days to record a book. My audio booth is the bathroom, as I live in a noisy area of town. I drag in a table and a soundproof box, then surround the area with doonas and eiderdowns. As a sole operator, I work the computer and equipment, record the voice, and then afterwards check the recording. Then come the re-takes, microphone ‘pops,’ and the occasional missed or wrongly pronounced word. I’d have to sell a lot of audiobooks to get my investment back.
I’ve been experimenting with AI voices and used several ones for PR videos that helped promote my books. The images were created in MidJourney. I sourced the voices from Eleven Labs and Narakeet, then did the animation in D-ID, fine tuned the audio in Audacity and lastly pulled it all together in Final Cut Pro. The process is a challenge to understand, but once you get it, it flows easily. To do this kind of work before required a huge amount of time and expense. But now I experiment. A whole new creative door has opened. Although I never let ChatGPT or its siblings write for me – that is my domain.
When the voice-over work dried up, I suspected my time in front of the microphone was at an end. But like most things in my life, when one door closes, another opens. I offered my voice to an AI audio company. I had been using Eleven Labs for some time, so I approached them.
To get my voice accepted, I had to upload up to three hours of my voice, preferably recorded on the same microphone and in the same studio. Fortunately, I have samples from many television documentaries I had recorded over the past few years. And the same microphone.
Destroying the Industry – A friend suggested I was destroying the voice-over industry by doing this. That I had had no voice work for a long time was not his concern. But I could see where he was coming from.
Previously, I mentioned I had no interest in recording audio for my novels. Nor am I interested in paying thousands of dollars for someone else to record them for me. But I am interested in having an AI voice, either mine or another, to record my books. The difference in price is mind-blowing. I could use my voice for a 400-page novel for about $US100, but it is not as easy as pushing a button. For example, is it read, read or read? (read, red or reed) ? At present, heteronyms make it a bit dizzy. So, a fair bit of pre-production work is needed.
I was wary of the quality until I listened and experimented. The quality is improving all the time, and Eleven Labs is introducing voices that can be incorporated into a wide range of apps and programs.
It’s Not the End of Voice Work – Of course, films, documentaries, and ad agencies will still use human voice-over talent. And human producers – some scripts are too complicated for AI (at the moment). Having worked both sides of the microphone, I am often amused when a producer tells me to do another read but to add ten percent more energy. I have never worked that one out. How can I pull ten percent out of a hat? In these situations, I do some different facial expressions that might indicate ten percent and end up delivering exactly the same read. Most often, they are happy with it. Amazing what a ten percent face pose does.
When you choose a voice with Eleven Labs, the originator of the voice gets a commission. A couple of cents per 200 words. Talent can also block their voices to certain restricted categories.
Here’s a link to Eleven Labs, where my voice is identified as ‘Patrick International.’ When I lived in Asia, many clients liked my voice, as I didn’t have an identifiable accent. Although AI has produced me sounding a tad British. That’s ok, as long as it’s not cockney.
Voice Models – Eleven Labs has four different AI Voice Models of each voice. To me my voice sounds different on each one. English Turbo 2 is close to my actual voice. Multilingual 1 offers me with a slight American accent and Multilingual 2 with a British accent. English 1 is neutral. Have a listen.
Four AI Voice Models
A Changing World – I suspect AI voices will change the world. We will find uses for AI voices in ways we never thought of. Travel and language will be at the top of the list.
Before AI, the world of voice-over was mysterious and expensive to most. Now it is here and can be accessed by anyone. Recently, I had my AI voice read a long magazine article. Then I sat back, read, and listened. Now, I have double the chance that my brain will understand the content.
I am old enough to remember when we didn’t have a phone, and I remember my first glimpse of the World Wide Web. It’s very exciting to be part of something new.
Patrick International is the name of my AI Voice
Disclaimer: I get a commission if anyone signs up to Eleven Labs through my link. I also earn a few cents for every hundred words if my voice is used. I earn lots more with my real voice.
