Last week iSpeech, a text-to-speech startup, announced that it had acquired its first publishing client: Pearson. Claiming "the most natural-sounding TTS audio on the market" and previously known for driving-direction applications and audio cues for the home, iSpeech wants to help publishers automate the creation of audiobooks. The default voice on iSpeech is pleasantly lilting, but it's still definitely not human, and more akin to a pre-recorded bus announcement in its odd pauses and stumbles. Pearson intends to use the service primarily with textbooks; the robotic voice is definitely not ready for the emotional range and styles of fiction yet.

Text-to-speech has been tried many times before, but it's technically very difficult to pull off, and legally complex. Amazon got into trouble in 2009 when it rolled out basic text-to-speech for the Kindle. Publishers said that Amazon was abusing its copyright and it withdrew, much to the chagrin of, among others, the blind and partially sighted community, reliant on audiobooks, who briefly gained access to Amazon's vast library. The problem is the cost of creating an audiobook. This involves studio hire and voice talent and can reach thousands of pounds, which is often difficult to justify in sales, while iSpeech offers automation costing just fractions of a penny per word.

But while the cost of creating quality audiobooks prevents much contemporary work being made available, older work is being "acoustically liberated" by collaborators on the internet. LibriVox, founded in 2005, recently passed 6,500 books in its collection: all in the public domain, all recorded by volunteers. LibriVox's moderators assemble teams of participants who take a chapter each. While the tone might sometimes vary as much as iSpeech's robot, LibriVox's enthusiastic volunteers are now putting out three classic audiobooks a day, a victory for networking the passion of readers.