AI voice cloning is a tool. It can be used, and it can be abused.
Last week I was made aware that Elecrow, an electronics manufacturer and distributor, was using an unauthorized clone of my voice in some of their YouTube tutorials.
I couldn't prove it at the time, but there are AI identity detection services like Resemble.ai that can at least give some input into whether it was a direct clone or just something that sounds 'midwestern US male.'
Anyway, I made a short video about it, because I wanted to make a point that it is not okay for a corporation to clone someone's voice and use it in their videos without that person's consent.
When I posted it, I saw a lot of hot takes. One common one was that 'AI is inevitable', and I should just accept that anything I post online is in the public domain and can be used by anyone, so stop complaining. Another one was that I have a generic voice and they probably just used a generic AI voice, so I shouldn't jump to conclusions.
The majority of people saw it for what it was: my voice being stolen and used for someone else's profit.
In the video, I did mention it was Elecrow, but I didn't call for a boycott, or ask for everyone to tar and feather them. There are a couple reasons for that.
First, I didn't have 100% proof they cloned my voice, and I hate the idea of 'guilty until proven innocent'. I even said "I'm hoping beyond all hope it was an honest mistake."
And second—this may surprise a few of you—I went to a Catholic seminary for a few years. Now obviously I'm not a priest, but my faith does influence how I live.
A big thing at the seminary was the idea of 'fraternal correction'. Basically, if you have a greivance with someone, take it up with them first, then if that gets you nowhere, go to the authorities.
Some people said I should lawyer up, or send takedown requests through YouTube.
Here's the problem: I don't usually enjoy thinking about AI, I really don't. I have a ton of projects on my desk, like right now I have a new time HAT for the Raspberry Pi and I want to tinker with it and share the project I'm working on.
When you get into anything legal, it sucks up a lot of time and money. My main goal was to make the point that it's not okay for a company like Elecrow to clone my voice and profit off it.
And besides, if you support my work on Patreon, I want that money to go to making cool videos to inspire people to do new things with computers and electronics. I don't want that money going to a lawyer.
Leaning into the drama would also feed into the justifiable anger I felt about the situation. That's another thing I learned in seminary, as a Belizian friend would say, "do not let the sun go down on your rat."
He meant to say "Don't let the sun go down on your wrath."
I have a wife and five kids, they have their own lives. For their sake, I don't have time to sit online stirring up drama. I really don't want to get hot and bothered about corporate ethics and all that. But believe me, I would, if Elecrow didn't respond. Because controlling anger is important, but that doesn't mean you have to be a pushover.
Luckily, Elecrow did respond.
Letter from Elecrow's CEO
Early Monday morning, the CEO of Elecrow, Richard Lee, responded to my email. He said:
Hi Jeff,
This Richard the CEO of Elecrow. At first please accept my sincerely apology for the possible infringement issues!
Now, really quick: Elecrow's a Chinese company. I couldn't write an email in Chinese without grammar issues, so I'm not going to worry about that at all. I care about the content, not the grammar. But he goes on:
I noticed your complaint this morning from my colleague, we take this matter very seriously and conducted a swift internal investigation immediately. Through investigation, we found that the video was made by one of our staff called [blank] who is a fresh graduate and not fully acquainted with our company culture and didn’t got enough training, he just thinking how to make the video get more popular but didn’t consider the copyright problem, also the video was released without the approval of the department manager, we are ashamed and incredibly sorry for make this case happen!
Now that's a good apology. I mean, whether or not everything is true, I can't judge. I'm willing to accept it though, and even if it wasn't a fresh graduate and the CEO himself made the decision to use my voice, just accepting responsibility is something that so few tech leaders do these days.
And the fact they got this email out early Monday morning, after less than 24 hours from my email, means they gave the situation the priority it deserves. I'll have some thoughts on calling out the junior employee later, but Richard goes on:
We will do following things immediately:
Video Removal: We have immediately removed all videos involving cloned voices upon receiving your email. This is a direct response to the error in our work, and I kindly ask you to monitor us accordingly.
They did take down all the videos I had mentioned, and Tom's Hardware also went through a few more videos and didn't hear my voice in any of them. So good work on that. And I'm glad they're asking for accountability. They didn't just say "we took down the videos, now stop bothering us!"
Moving on:
Internal training: we have provided a profound criticism to our marketing team, also we will strengthen internal management and personnel training in the future, to enhancing their awareness of copyright issues and professional ethics.
Honestly, as someone who's made my share of blunders—one time bringing down a multi-million dollar product launch for a few hours because I used a new AWS service without monitoring a critical resource limit... I hope they take this as a learning opportunity. It's a good time to reflect on not just voice cloning, but any other bad marketing tactics, like scraping email lists for marketing.
Compensation: we willing to compensate you for any losses caused by this incident.
I spend a lot of time thinking about sponsorships, vendor relations... all that stuff. The number one guiding principle for my writing is honesty. I don't ever want someone to think any word I said is not my own. That's why I don't do full sponsored videos, where I follow a company's talking points about a product. Nor do I have sponsored posts on this blog. And that's why I don't sell my voice for use by others.
But compensation... I mean, I have had to delay publishing my next video, and content is how I earn a living. And I did ask my lawyer for some advice—lawyers ain't cheap! And, behind-the-scenes, I talk to a lot of vendors, if one of them saw one of Elecrow's videos and thought 'maybe we don't need to talk to Jeff because he's endorsing Elecrow?'...
What I've proposed is they send me $2,500, and I'll donate $2,000 of that to the UOAA to kick off Ostomy Awareness Day in the US. I have an ostomy, and I want the UOAA to help other people who might not have the same opportunities I do. In fact, let's make some lemonade out of life's lemons today: if you want to support ostomates in the US, please consider joining me in donating!
But back to the email, Richard goes on:
Preventing Future Infringements: The occurrence of this incident has made me deeply aware that our content review mechanism is not comprehensive enough. We will strengthen our content review and release processes, and welcome your active supervision. We guarantee such kind of issue will never happen again.
That is good to hear. He finished his email with a couple more remarks about how Elecrow respects IP rights and again apologizes about the whole thing.
He also asked if I'd remove the original video, but I don't like to take things down if I think they're truthful and not harmful. So for now, I updated my pinned comment on that video, and I'll link to my new video in the description.
Judging their response
Now overall, I'm happy they responded at all. So many companies these days address controversy by not addressing it, and sadly, that often works. On the flip side, some could construe the response as a response to being 'caught red handed'. But either way, responding quickly, and with a fitting apology and steps to improve is better than 99% of the tech world, so I can't complain about that.
The greatest irony would be if they used AI to generate the response (but I don't think they did).
One thing I will complain about a little is the calling out of the individual who they say made these videos. As someone who worked as a lead on a number of teams, but never in management, I saw firsthand how people can grow from their mistakes. Especially junior employees who make big ones.
You have to have discipline for mistakes, but good discipline leads to growth. There are times where you fire someone for a mistake, but this isn't one of them.
The idea of trying to get more views on a tutorial series by using a recognized voice is a good one. Just the means of getting that voice was wrong. That's the lesson to be learned.
I asked about it. Richard clarified that the employee won't be fired or blamed, he said the company and the process was the problem, not the new employee. That was very reassuring.
How they did it
All this is about Elecrow's response, but what I haven't covered is how they did it. How was it so easy to clone my voice that a new college grad could do it for a bunch of tutorials on their channel.
Well, tools like ElevenLabs make it shockingly easy. I paid $5 and I immediately had access to a tool called "Instant Voice Cloning".
I just grabbed one minute of my talking from my latest YouTube video, uploaded it, and bingo, I have "AI Jeff".
Some creators are even using this for their own videos, whether it's for shorts, or secondary videos where they don't have time to record a-roll.
If you're unscrupulous, you could steal another creator's voice just by clipping part of their video!
Like I wanted to ask Dr. Ian Cutress, from TechTechPotato, if I could use his voice.
See how this could be dangerous?
And one reason this is getting so popular is because hiring a professional voiceover artist for just a few hours of video costs hundreds or thousands of dollars and requires a relationship. Hiring ElevenLabs and using AI? $99 gets you 8+ hours of text to speech a month.
How you can protect your own voice
In the video I have that goes along with this post, I had a conversation with the co-founder of Resemble.ai, a startup that's tackling issues like AI voice identity detection with new tools. You can watch that interview starting around 9 minutes in:
Conclusion
I'm a firm believer in making lemonade out of lemons. And honestly, if Elecrow improves their process, and I don't have to worry about them stealing my voice, that's at least a little win.
And I found out YouTube does allow creators to take down videos using unauthorized AI clones. Thanks to Rene, the YouTube Creator Liason for pointing that out.
AI is a tool. And right now it seems like it's being used like a flamethrower a lot more than a pocketknife.
It can be used well. I hope people will do that, and not use it for harm. One way to ensure that is holding people and companies accountable for improper AI use.
Comments
It's really nice to see that someone is actually living the idea of loving your neighbour and publicly committing to it! I try to do the same, even if it's increasingly difficult to see the good in others these days. Perhaps it is even the only way to find our way back to each other. Confrontation and lawyers are less likely to help. Keep it up! (by the way, also with your content, really like it!).
Nice post. I like that you broke down your emotional play by play. Guilt. Blame. Injustice. Etc. As well as your practical play by play: request an explanation and apology, consider the outcomes, mull over the results.
I will say though, we are in a strange state of the world, where we no longer feel like we can enforce laws, or utilize our rights. Laws were made to protect people, but instead they protect corporations, I suppose we threw that baby out with the bathwater, when we let corporations become people.