Apple’s iOS 17 will evidently contain a new feature that can replicate your voice in 15 minutes by reading a random paragraph, and then allowing the phone software to effectively sample your voice, and use AI to replicate it.
Whilst I applaud the overall idea in the sense that it could assist those who suffer from ALS in the future by allowing them to convert typing into their own voice, I also see a note sinister side where this technology could easily be abused.
As a starting point, think “my voice is my password”… Then there’s always the possibility of being leveraged for deep fake purposes without consent. And what security controls are in place to protect that data, which easily falls into the personally identified category?
Liife enriching and life changing, yes, but the changes attract a new array of nefarious possibilities. Clearly, it’s possible to protect the voice synthesis with a biometric footprint, but how simple and cost effective works this be to implement?
Would love to hear opinions on this topic.
Author of OGProxy, Swatch, Threaded, Card, and most of Sudonix's bugs…
Designing and creating tools for a purpose is one thing, but whether people will use them for this purpose is another thing. It might be useful for a certain small groups of people, but overall I believe people would prefer exchanging voice recordings over this.
Many voice characteristics are definitely changing over time, so I suspect this will be useful for “security” purposes. Well, I guess it can definitely prevent some very obvious scams, but overall it will always be a secondary tool that “might” be used.
@phenomlab said in A voice that sounds just like yours in 15 minutes?:
Whilst I applaud the overall idea in the sense that it could assist those who suffer from ALS in the future by allowing them to convert typing into their own voice
This reminds me this project:
I guess everyone is familiar with Stephen Hawking’s voice. This project was created so people can choose a synthetic voice that fits their age, gender, and personality. I remember reading some sentences and donating “my voice” 7-8 years ago… but I do not know if my voice was ever picked 
I am definitely not a Freddie Mercury or George Michael 
@crazycells great points raised here. It’d be interesting to see how close AI can get to your own voice with the least amount of effort, or the ability to provide a good match based on one sentence alone.
As I’ve already stated, my concerns initially around security with be anything that can distinguish your voice. We’ve already seen what deep fake technology is capable of, and this (to me at least) is a cut down “affordable” version of that built right into your phone.
