In November, NBE editions published the first “ meta-book», available in bookstores. In parallel with the paper version, a collection of NFTs has been designed by the Artificial Intelligence of the startup Selas AI. Each reveals a snippet of the work, or the meeting of publishing with Web 4.0.
Two years ago, Google Play Books unveiled its automated reading solution: synthetic narrators, to shape audio books without resorting to actresses or actors. This December 2022, the program has been generalizedin beta version: the publishers of create the audio version of the published works for free.
While many tech companies, like TikTok, are already relying on the increasingly sophisticated capabilities what about the publishing world and enriching its range of audiobooks? The big houses and platforms rely on famous readers, like Spotify : in 2020, the platform shared an audio version of the first volume of the literary series Harry Potter, read by actor Daniel Radcliffe himself.
For Amazon subsidiary Audible, filmmaker and playwright Sam Mendes adapted three works by Charles Dickensbegining withOlivier Twist . The series is recorded in the London studios, a short walk from the neighborhood described by Dickens in the novel. The world’s largest distributor and creator of audio titles, Audible currently only distributes audiobooks told by humans.
A question then arises: what does Denis Podalydès, narrator of flesh and blood, bring to the exercise of reading?
AI and lower costs
In the audiobook, editorial projects require a certain investment. According to Mark Piesing, the use of artificial voice can reduce costs, for standard-length fiction, by $2,500 to $400. Thus, speech synthesis technologies (TTS) would also bring democratization production of audio titles. On the other hand, the latter confesses that the artificial voices are improving, but still remain unequal depending on the language.
Taylan Kamis, co-founder and CEO of DeepZen, the company credited with delivering the world’s first AI-powered audiobook in 2021, explains:Over 100 million books in print worldwide and 20-25 million e-books, but only half a million audio works, 90% of them in English and half produced in the past four to five years. »
And to add:The cost and time required to produce an audio work constitute the main obstacle, especially when it comes to non-English speaking markets. For a publisher, the opportunity is vast.Today, computer models such as WaveNet, DeepMind, and Tacotron spawn neural networks that mimic the functioning of the human brain to continuously improve the speed and accuracy of their AIs.
Each start-up relies on different models, as a basis on which to build their own applications, allowing them to stand out in this market of the future. DeepZen wishes for example “capture the emotion» for its text-to-speech technology.
What should emerge is mass-produced, low-cost, AI-powered productions, and in parallel, high-quality human-voiced productions, and more expensive. Moreover, a work that one listens to for its information will require less inflection of the voice than the prose of Louis Ferdinand Céline.
According to Mark Piesing, “publishers’ attitudes will continue to evolve as technology improves and Gen Z grows. Over the next five years, rising demand for audio content, the cost of producing voice-over audio books, and the desire for democratization will shatter opposition“.
