NEW YORK: As people prepare for the disruptive impact of artificial intelligence on their work and daily lives, those in the audiobook world say their field is already being transformed.
AI has the ability to create human-like recordings at line-up speed while bypassing at least some of the services of human professionals who have long made their living from voice.
Many of those companies have already experienced a sharp decline in business.
Tanya Eby has been working as a full-time voice actor and professional narrator for 20 years. She has a recording studio in her home.
But over the past six months, her workload has halved. Her reservation is currently until June, but in normal years she will be extended until August.
Many of her colleagues report similar declines.
“It seems natural that AI is affecting us all,” she told AFP, although other factors may be at play.
While there is no label that identifies AI-assisted recordings as such, experts say thousands of audiobooks in circulation today use “voices” generated from databanks.
Our most advanced service, DeepZen, offers rates that can cut audiobook production costs to less than a quarter of the cost of traditional projects.
The small London-based company uses a database created by recording the voices of several actors who were asked to speak in various emotional expressions.
“For any audio we use, we sign a license agreement and pay a recording fee,” said DeepZen CEO Khamis Tyran.
For all projects, “we pay royalties according to the work,” he added.
Not everyone respects that standard, Eby said.
“There are new companies popping up that aren’t so ethical, and some of them are using the voices they have in their databases for free,” she said.
Tyran acknowledged that there are “gray areas” being exploited on some platforms.
“They’re just combining your voice, mine and five other voices to create another voice… say it’s nobody’s.”
All audiobook companies contacted by AFP denied using such practices.
Texas-based startup Speechiki uses both its own recordings and voices from existing databanks, said CEO Dima Abramov.
But that will only happen after a contract including usage rights has been signed, he said.
Future of coexistence?
Five major U.S. publishers did not respond to requests for comment.
But some traditional publishers are already using so-called generative AI, which can create text, images, video and audio from existing content without human intervention, according to experts contacted by AFP. It says.
“Professional narration has been and will continue to be a core part of the Audible listening experience,” said a spokesperson for the Amazon subsidiary of the American audiobook giant.
“But as text-to-speech technology improves, we see a future where human performance and text-generated content can coexist.”
US tech giants deeply involved in the explosive growth of AI are all pursuing the promising business of digitally narrated audiobooks.
“Anyone can access it”
Earlier this year, Apple announced a move to AI-narrated audiobooks, which the company said would make “audiobook creation more accessible to everyone, especially independent authors and small publishers.” rice field.
Google offers a similar service, which it calls “automatic narration.”
“We need to democratize the publishing industry because only the most famous and big names are on audio,” Tyran said.
“Synthetic narration has only opened the door to old books that have never been recorded, and all future books that will never be recorded for economic reasons,” added Speechiki’s Abramov.
Considering the cost of human recording, only about 5% of all books are converted to audiobooks, he added.
But Abramov argued that the growing market would also benefit voice actors.
“They will make more money and record more,” he said.
the human element
“The essence of storytelling is to teach humanity what it means to be human, and we strongly feel that we should not give it to machines to teach us what it means to be human,” says the actor. said Emily Ellett, audiobook narrator and co-founder of the Professional Audiobook Narrators Association (PANA).
“Storytelling should be completely human,” she added.
Eby highlighted frequent criticisms of digitally generated recordings.
Compared to human recordings, AI products “lack the emotional connection,” she said.
But Eby said he fears people will get used to the machine-generated version. “I think that’s what’s happening quietly,” he said.
Her wish is simply, “I want companies to let listeners know they’re listening to AI-generated work…I just want people to be honest about it.”
