Text to speech has a long history, with champions and critics along the way. With a massive technology breakthrough in 2018, the voice over game has changed. The TTS of today, such as the lifelike voices created by WellSaid Labs, is miles away from what you may have heard before 2018. Synthetic voice is quickly becoming a competitive option for eLearning voice narration.
We’ve written at length about the benefits of text to speech: it’s high-quality, affordable, and easy to use. While many eLearning developers use it to provide content narration, most haven’t yet made the switch to higher-quality, human-like services.
To help you see what you can achieve with new TTS, we’d like to introduce you to WellSaid Studio. One of the latest TTS offerings, Studio is built on the most recent AI voice research, and it shows. In evaluations conducted by a third party, WellSaid’s synthetic voices rate nearly on par with the human voices they were built on.
Using this sample eLearning script produced by voices.com, we decided to audition two voices from the WellSaid Studio library to demonstrate their capability.
Here’s our sample script:
Up to now, you’ve learned how to identify workplace hazards and how to report them to your supervisor. In this module, we’ll discuss the hierarchy of control measures that can be used to eliminate hazards from the work process altogether.
Audition 1: Wade C.
Audition 2: Sofia H.
Saying it well, just in time
From its 1939 unveiling by Bell Labs at the World’s Fair to now, artificial speech has developed in fascinating ways. Computerized speech saw a great improvement when researchers began mapping phonetic patterns and applying those learnings to research data, a technique called “Speech Synthesis Markup Language” or SSML. Amazon’s Alexa, for instance, is built using SSML.
WellSaid emerged right as the technology made a critical shift from these older methods to deep learning. Michael Petrochuk, co-founder and CTO, describes how WellSaid Labs found its beginnings in the middle of this technological breakthrough:
“There was a lot of great and new research coming out during the time that we started the company. Some of the research we based our work off of was only a month old.”
Tacotron 2, “one of the first papers to claim to achieve human parity in TTS”, came out in February 2018. One month later, WellSaid Labs was conceptualized.
High-tech for all
This technology’s incredible capabilities remain the primary driving force behind the growth of WellSaid Labs.
“We wanted to give people access to this technology. We knew that it could be really helpful for people working with voice over to create all sorts of content. If we really had human parity technology and if we could open it up, then it would be much easier to work with voice over, iterate with voice over, create with voice over, add voice over to content creation projects.”
While plenty of content creators contentedly use older text to speech synthesizers, many others dislike the relatively flat results. Some critics even argue that the lack of emphasis in AI voices makes learning impossible. WellSaid gives people the opportunity to create content with life-like voices, emphasis and all.
AI for good
TTS enables course developers to make existing courses accessible, but it also does more: It’s a powerful tool with promising applications. TTS can change lives.
“Creating content is not where this technology stops,” says Petrochuk. “If we fully realize the company’s vision, there are a lot of people we can help, a lot of use cases.”
One potential use case is voice preservation. “We know that there are millions of people that lose their voice or have already lost their voices,” Petrochuk continues. We now have the ability to give someone a synthetic version of their own voice. As long as we can record their voices before they’re lost, or use existing recordings, the technology to create an AI voice for them is available and accessible. Even without the option to craft a customized voice, human-like text to speech is a huge improvement over the current status quo. Giving the voiceless access to a human-sounding text to speech service can make a world of difference.
“For those people that want to use a text to speech service, we could provide one that is much more life-like, much more expressive, one that can give them a lot of freedom in how they express themselves. And that is certainly something that this technology enables.”