As we wait out that curious stage of artificial intelligence (AI), unsure whether it will precipitate the end of civilisation or have no greater impact on the global economy than the fax machine, there is one preliminary conclusion. AI narration is surprisingly good, and sometimes better than humans.
After years of attempting to outwit The Atlantic’s paywall, I recently subscribed to the magazine and downloaded the app. To my initial disappointment, the narrated articles were read by an AI voice. For so long, these have been, well, dreadful. Imagine asking a small child who has just completed their fourth or fifth Biff, Chip and Kipper book to read aloud a George Packer polemic on what the working class of America really wants. So it was with some trepidation that I pressed play on the first article.
Speech-to-text software has been around for some time, but this hits different. Produced by ElevenLabs and News Over Audio, the tone of voice undulates naturally, the pronunciation is pretty much correct. There is the odd error – for example, it got tripped up over the term ‘op-ed’, confusing it for the past tense. But other than that, I quickly forgot I was listening to a non-sentient being.
Indeed, the AI voice is more convincing than half the audiobooks I listen to. My heart sinks when I see that a book I want to download is narrated by the author. Not everyone sounds as dreamy as Robert Caro or Stephen Kotkin. Just because you wrote it does not mean you are the best person to read it. If your novel has characters with different accents, and you can’t perform them to any degree of accuracy, please do not attempt to do so.
The Economist has for as long as I can remember produced a weekly audio edition read by a stable of professionals. I know each of their voices like the Test Match Special team and am hit for six when a new one starts talking at me about the accuracy of Argentinian inflation figures.
But let’s face it, not every media organisation has the time or financial resources to offer that as a feature. And there are so many books that I would like to listen to (I’ve written previously about my adulthood-induced inability to actually read) that would only be available through AI narration.
Like any technology, AI will be used for evil as well as good. Beyond nuclear risk there are deepfake scams, carbon emissions and use in propaganda. Soon, we may not be able to trust anything we hear. But if AI can make every book available for download, at least we will be entertained as the world comes to an end.
