With the increasing availability of Artificial Intelligence (AI), companies are now selling AI-created products such as images, database searching tools, chat bots, and even AI-narrated audiobooks. AI models are created by first designing an algorithm, and then feeding the algorithm content that trains it to then create similar content. Companies have been able to create AI models that are considered “synthetic voices” by contributing voice data to the model. The AI model then replicates human voice in reaction to text prompts. This is the manner in which AI-narrated audiobooks may be produced in minutes.
Audiobooks are expensive to make, but the cost of using a synthetic voice to narrate an audiobook can be ten times cheaper than recording an audiobook in a studio, according to the synthetic voice vendor Speechki.
Uploading a PDF of a manuscript can result in a fully narrated audiobook in as little as fifteen minutes. However, this does not mean that the audiobook will be in final, listenable form immediately. Although synthetic voices are convincing replications of human voice and speech patterns, there is potential for the AI model to misinterpret the structure of a sentence, or perhaps pronounce a word incorrectly, particularly if the word belongs to a language other than English. For this reason, it is important that any publisher producing audiobooks using synthetic voices either employ the services of a synthetic voice vendor, which will provide editorial services along with the synthetic voice, (such as DeepZen), or be prepared to use the vendor’s client portal to edit the synthetic voice reading of the book themself. Instead of calling a narrator in to record audio and then re-record when errors are discovered, editors will be able to edit the pronunciation of words, as well as speed of delivery and tone of voice.
Google Play Books began offering “auto-narration” services to publishers or authors who have distributed e-books through Google Play in the summer of 2022. Since then, it has offered the service for free: first as a beta test, and now for an undisclosed limited time.
Any auto-narration created by Google Play Books can then only be distributed through Google Play Books. In other words, the narration cannot be uploaded or distributed elsewhere.
The controls that Google Play Books offers to edit the auto-narration (AI narration) include changing the voice, language, and speed of the narration; the ability to add additional characters to the narration, which could come in handy for character voices throughout a manuscript; and editing pronunciation of words by choosing from suggested pronunciation or by entering phonetic text manually or recording yourself speaking the word. Edits to pronunciation of a word can be applied throughout the manuscript, or to only one instance of the word.
Google Play Books also offers tips for preparing a text to be auto-narrated. Suggestions like remove footnotes in the case that they are not removed automatically, spell out uncommon abbreviations, and replace censored words that may use special characters for visual representation with an audible representation like “blank” (Google Play Books).
These controls will allow publishers to create audiobooks for less money, but publishers shouldn’t expect readers to choose AI narrated audiobooks over human narrated audiobooks every time. The nonfiction genre is especially suited for AI narration due to the narration delivery of nonfiction requiring less emotion compared to fiction.
Although not every book makes a great candidate for AI narration, hopefully more audiobooks will be produced with the use of a cost-effective tool for adapting manuscripts into a listenable format.
Written by Paige Brayton.