This AI Startup Wants You to Read Audiobooks to Yourself

AI startup Speechify is putting its own spin on audiobooks and giving you, the listener, a leading role. You get to be the star if you want.

You can import your own voice to make an AI clone and then listen to text with your voice or your girlfriend’s, as in the case of CEO Cliff Weitzman.

You can also choose from celebrities like Snoop Dogg and Gwyneth Paltrow, who have signed on to add their voices as options. The twist being these are AI generated, not the celebs themselves reading.

“You can just pick your own voices and that’s a great experience,” Weitzman said.

It’s this ability to choose whatever voice you want and to turn any book into an audiobook that Weitzman argued sets Speechify apart from titans of industry like Apple Books, Audible and Spotify.

Audiobooks are a hot commodity. According to the Audio Publishers Association, 2023 marked the 12th consecutive year of sales growth, with a total of about $2 billion for the year. The APA also found 52% of US adults have listened to audiobooks at some point, which is equivalent to about 150 million people.   

With artificial intelligence, and especially generative AI, exerting its influence far and wide, we’re seeing entrepreneurs seek to harness the technology to challenge the status quo in a variety of industries, from law to medicine and even generative AI itself. Seven-year-old Speechify is positioning its text-to-speech reading tool as an alternative to traditional audiobooks through the use of AI-generated human voices. 

As a child with dyslexia, Weitzman relied on his parents to read books aloud to him. But when he got to college, he couldn’t find audio versions of his textbooks, so he built a program to read to him using deep learning, an AI technique that teaches computers to process data like the human brain does, and what’s known as concatenative text-to-speech, a form of speech generation that taps into pre-recorded samples of speech.

The native Hebrew speaker also included the ability to change the speed — a feature Speechify retains today.

“When I started out, I didn’t speak English, so I would listen to everything at 0.75x speed and then with time I increased to 1x,1.25x, 1.5x, 2x, 3x,” he said. “If a sentence was easy to understand, I’d make it really fast. If the sentence was hard, I’d make it really slow.”

Weitzman’s brother Tyler joined as a co-founder in 2018 and has served as head of AI and president since 2022. Tyler Weitzman helped develop the algorithm that eventually became the first version of Speechify. It was trained on 100,000 hours of audio so the reading voice sounded human.As the product improved, the startup signed partnerships with celebrities to use their voices as well.