Tell Me a Story, Robot

February 11, 2020

Make it a story of great distances and starlight…”– Robert Penn Warren from Tell Me A Story

If you’re reading this, then you’re a reader, That’s obvious, even tautological (nice word!). What I really mean is that you’re most likely a lover of books and good stories. But when someone says, “Tell Me A Story,” they are not talking about reading and writing. They’re talking about speaking and listening. Telling stories out loud is more primal. It runs in our DNA, and it was really the way stories were enjoyed and passed down up until the time of the Bible and other foundational texts like the Iliad and Gilgamesh. Even then, very few people could afford those painstakingly, handwritten manuscripts, so they were recited – enter the dawn of the audiobook. Read on and join me on a journey into Text-To-Speech(TTS) and the burgeoning tech effort to turn robots into storytellers.

Bedtime Stories

As parents, we have dutifully passed down the telling or reading of stories to our kids at bedtime. The tales delight and calm them, planting the seeds of pleasant dreams. Maybe this explains the popularity of podcasts and audiobooks. We have a relatively new platform for the dissemination of the spoken word with the surprise and delight of stories well told.

I grew up addicted to the spoken story with Mystery Theater on the radio. My parents’ generation listened to shows like The Shadow and Our Miss Brooks. And almost everyone has heard the famous Orson Welles’ War of the Worlds. As an adult, my love of audiobooks dates back to Books on Tape that started up in the 1970s – BOT is still around. On my long work commutes, great stories whisked me away to distant places and triggered my imagination, shrinking my perceived commute time to nothing.

Audiobooks, Podcasts – Give Me More

Now that I am a writer, of course I would make my novels into audiobooks – NOT SO DEAD audio appeared in early 2018 and NOT SO GONE earlier this year. That journey began with interviewing ten narrators/actors to do the fourteen different characters. Since by now my wife, Amy, and I listened to hundreds of audiobooks in the years since Books on Tape, we had a good ear for what makes a great narrator. You don’t want an overly dramatic narrator – that would take away from the story – but you do want an actor who knows when to speed up with the action and emphasize specific words. And you don’t need a man who can sound perfectly female when he is performing a woman’s part – only slight changes in pitch and cadence will do the trick. The narrator is really a midwife who helps readers give birth to their own personal experiences of the story.

Charles Levin Author

Text-To-Speech – Is It Any Good?

With this in mind and being kind of a techy myself, I was intrigued when invited to beta test a voice synthesizing program from the well-funded tech startup WellSaid. Text-to-speech (TTS) is a rapidly-growing field of machine learning with many players from start-ups to the likes of Google, vying for the lion’s share of the projected $3 billion TTS market. Could TTS replace human narrators at much less cost? Would the quality be there?

WellSaid’s YouTube demo was impressive. So as a beta tester, I had the opportunity to try it myself. I inputted my text, selected from one of three digital narrators, and a human-like voice read it aloud. Here are the first few sentences of the first chapter of NOT SO DEAD done by the Voice Synthesizer – compare them to my human narrator,  HERE. What do you think (please leave a comment below)? I also tried Google’s TTS which boasts of 100 voices to choose from in more than 20 languages. You can try your own Google TTS experiment for free here.  My opinion: for a straight reading of the news or maybe even for advertising voice-overs, it’s fine. The words are pronounced correctly, but TTS has a way to go for storytelling. Human audiobook narrators are safe for now. Storytelling requires a great deal of nuance in the tone and emphasis to deliver an exceptional audio experience. And for the foreseeable future, that appears to be way beyond the current software wizardry.

In fact, it took decades beyond landing a man on the moon to reach 99% accuracy with Speech-To-Text (the flip-side of TTS) aka turning your voice dictation into the written word. The futurists had no idea that the task was more complex than a moonshot, but Speech-To-Text eventually got there. Take this a step further and maybe we writers and creators need to look over our shoulders as Artificial Intelligence can actually write and create fiction and do it well. So well, that when Open AI created a very powerful content creation program, it scared them and they shut it down. Similarly, who thought computers could beat the world’s top chess champion or the top Go master and they did. What does all this mean for our future? I’m not sure, but I will be watching and listening and still writing with anticipation. In the meantime, I don’t believe anything will replace Mom or Dad reading you Good Night Moon before bed anytime soon.

Happy reading! Or should I say listening…

— END —

%d bloggers like this: