2021 Speech Industry Award Winner: ReadSpeaker Gives a New Voice to TTS
ReadSpeaker has been leading the way in text-to-speech (TTS) for more than two decades, delivering high-quality, lifelike voices for artificial intelligence-driven conversational applications, interactive voice response systems, and embedded applications. To date, the company, which operates as a subsidiary of the Memory Disk Division of HOYA, provides roughly 100 voices for dozens of languages and has created more than 10,000 digital voice interfaces for companies around the world.
And this past year, its TTS technology has started appearing in some of the most impressive spots. Japan’s Sanyo Electric Railway, for example, in February began using ReadSpeaker as the voice of its train announcement system, broadcasting throughout the trains in both English and Japanese.
Il Sole 24 Ore, one of Italy’s highest-circulation newspapers, chose an Italian text-to-speech voice from ReadSpeaker to read content out loud on its premium app. The app also features a voice-management system that enables users to interact with the app using their voices. They can open full editions, articles, podcasts, and other content from the publication’s archives.
And Honda’s new electric vehicle, the Honda e, will use ReadSpeaker’s TTS engine as the voice of the Honda Personal Assistant, an in-car interface that allows drivers to access real-time information about parking and charging stations, service availability, amenities, and much more.
In January, sComm, a developer of speech-generating devices for the deaf and hard of hearing, selected ReadSpeaker’s TTS to power the speech output from its UbiDuo devices.
“The partnership with ReadSpeaker has allowed us to take our UbiDuo SGDs to the next level,” said Jason Curry, CEO of sComm, in a statement. “With ReadSpeaker’s technology embedded into our devices, we are now confident that sComm delivers the best stand-alone SGDs in the world.”
ReadSpeaker is also providing the voice for Mixi’s Romi, an autonomous conversational robot that uses artificial intelligence to think of words to say in response to the content and flow of a conversation.
Romi has a unique voice synthesizer created using ReadSpeaker’s Custom Voice service, which creates a synthetic version of a real human narrator’s voice. In this particular case, it involved noted Japanese voice actress Yoko Honna.
European online retailer bol.com also turned to ReadSpeaker to launch a Dutch custom branded voice for its Google Assistant. The voice answers customer queries and delivers daily offers and updates.
“At bol.com, we are constantly focusing on what we can do better for our customers. Experimenting with innovations in voice helps us to better understand what our customers prefer and how we can make their lives a little bit easier. We chose ReadSpeaker to help us with that for the humanlike quality of its custom voices, its expertise, and its customer-oriented approach,” said Vera Rensink, business developer at bol.com, in a statement.
And ReadSpeaker partnered with SoundHound to include its TTS on the Houndify Voice AI platform. Developers using SoundHound’s Houndify can now add ReadSpeaker’s voices to their custom voice assistants.
“Sonic branding is the next big thing in digital voice, and with our TTS technology, we have the ability to help brands develop voices that can help bring their brands to life,” said Matt Muldoon, ReadSpeaker’s North American president, in a statement. “Our partnership with Houndify highlights the importance of completely personalized digital voice solutions.”
And while such customer wins are impressive, ReadSpeaker continues to update its products to keep them at the forefront of the industry. Its largest innovation this year was the introduction in November of Prosody Transfer within the ReadSpeaker neural TTS engine to enable even more lifelike TTS voices.
Prosody Transfer enables ReadSpeaker to transplant prosody models from one ReadSpeaker deep neural network voice to another. In other words, the company can now take the prosodic style of a source voice and transfer it to a target voice, leaving the latter’s voice color unchanged.
ReadSpeaker’s Neural TTS Engine is built on its foundational voice enablement platform, which today provides 90 TTS voices, available in 40 languages.
The company in July further expanded the number of languages available in its neural TTS voice portfolio, adding Arab, Indian, Norwegian, and Flemish voices.