Synthetic voices want to take over audiobooks

as a voice actor Heath Miller is sitting in his Maine converted boatshed studio recording the narration of the new audiobook, and he has read the text at least once. In order to provide his best performances, he recorded each character and any hints on how they should sound.Over the past two years, the role of audiobooks, such as telling the popular fantasy series man fighting monster, has become Miller’s main source of work.But in December, he was see a tweet Jon Richter, a British science fiction writer, has revealed that his latest audiobook doesn’t need the kind of artistry Miller provides: it’s narrated by synthetic voice.

Richter’s book listing on Amazon’s Audible refers to the voice as “Nicholas Smith,” but doesn’t reveal that it’s not human.To Miller’s surprise, he found that “Smith” had a total of about half a dozen voices from multiple publishers on the site — violations audible rules Says that audiobooks “must be narrated by people.” Although “Smith” sounds more expressive than a typical synthetic voice, to Miller’s ears it’s clearly artificial and provides a worse experience than a human narrator. It made giveaway mistakes, such as pronouncing Covid as “kah-viid” when referring to the pandemic.

Miller found ‘Smith’ – voice with sample posting San Francisco startup Speechki provides SoundCloud with more than 300 synthesized voices for audiobook publishing in 77 dialects and languages.He and other narrators and audiophiles discussed Online Human Audiobooks reported the titles to Audible, which eventually removed them. Although not many, the discovery that synthetic sounds were enough for some publishers to use them prompted Miller to start thinking about the future of his art and income. “It’s kind of scary because it’s my livelihood and the livelihood of so many people I respect,” he said.

Richter said he chose the artificial voice because the concept sounded right with its “uncanny valley” his book, which features a smart piece of software that doesn’t know Audible’s policies. “It was never my intention to upset or offend anyone,” he said. Speechki said it recommends that publishers determine that the narration is synthetic and inform them of Audible’s policy. Audible senior director Will Farrell-Green said in an emailed statement that the company uses automated and manual processes to enforce its rules, but “due to the volume of content on our service, non-compliant headlines do pop up from time to time. Audible’s “human-only” policy dates back to at least 2014, when synthetic voices were far less persuasive, and the company says the rule helps give listeners the performance they expect.

Synthetic voices have become less harsh in recent years, in part due to artificial intelligence research by companies like Google and Amazon, which are racing to offer virtual assistants and cloud services with smoother artificial tones.These advancements have also been used to make reality deception “deep fake. “Speechki is one of several startups developing speech synthesis for audiobooks. It uses in-house software to analyze the text to flag how different words are inflected, utters it using technology from cloud providers such as Amazon, Microsoft and Google, and interprets it. Hire evidence listeners to check for errors. Google is testing its own “Auto-narrative” service Publishers can generate English audiobooks for free using over 20 different synthetic voices.Audiobooks published through the program include Academic History of Drama and the novelist’s exploration cultural attitudes towards sex. Google spokesman Dan Jackson said its automated commentary supplements rather than replaces professional commentaries. “Our goal with Autonarrative is to create low-cost audiobooks for any ebook title and increase the accessibility of content for those who can’t read through ebooks,” he said.

Some publishers see synthetic voices as a way to meet growing demand for audiobooks, a healthier segment than the rest of the book business. Total revenue for U.S. book publishers fell slightly between 2015 and 2020, with ebook revenue shrinking but audiobook revenue surging 157%, according to the American Publishers Association. Consumers are increasingly comfortable with the format, aided by technological improvements in mobile apps, smart speakers and wireless headphones. But because of the cost of narration and audio production, most titles never make it to audiobooks, especially at smaller publishers, said Brian Carroll, rights manager at Indiana University Press.

IU Press licensed parts of its catalog for traditional audio production, but is now a Speechki client. It plans to release its first comprehensively narrated audiobook later this year. “All the other books now finally have a chance to be audiobooks,” Carroll said.

So far, Speechki’s technique has been impressive in tests, navigating the academic language of paleontology and philosophy titles, Carroll said.One book chosen for production is 80 Toasts Around the World, where the software must process text interspersed with words in other languages. “We think if it can do this, it can do anything, and it does it very well,” Carroll said.