(Google’s “standard” text-to-speech voices use a variation of this technology.) It offers more control over speech output, but has a muffled, robotic sound. Another technique, known as parametric TTS, doesn’t use recorded speech, but rather starts with statistical models of individual speech sounds that can be assembled into a sequence of words and sentences and processed through a speech synthesizer called a vocoder. This approach generates intelligible audio with an authentic-sounding timbre, but it requires many hours of recorded voice data and lots of fine-tuning-and its reliance on a recorded library of sounds makes it difficult to modify voices. Imagine a ransom note, where letters are cut up and pasted back together to form new sentences. In 2017, VentureBeat wrote: “Alexa is pretty smart, but no matter what the AI-powered assistant talks about, there’s no getting around its relatively flat and monotone voice.”Įarly versions of Alexa used a version of “concatenative” text-to speech, which works by compiling a large library of speech fragments recorded from a single speaker that can be recombined to produce complete words and sounds. But the early Alexa’s conversational style left a lot to be desired. The original Alexa, released with the Echo device in November 2014, is believed to have been based on the voice of Nina Rolle, a voiceover artist based in Boulder (something neither Amazon nor Rolle ever confirmed), and relied on technology developed by Polish text-to-speech company Ivona, acquired by Amazon in 2013. Whether or not you like the idea of “Grandma Alexa,” the demo highlights how quickly AI has impacted text-to-speech, and suggests that convincingly human fake voices could be a lot closer than we think. “Everyone thinks of speech recognition as the hot thing, but ultimately if you’re talking to something, don’t you want it to just talk back to you? To the extent that that can be personalized-with your voice or the voice of somebody you want to hear-all the better.” Providing accessibility for people with vision challenges, limited motor function, and other cognitive issues is another factor driving development of voice-tech, notably for e-learning. “There are going to be a lot more of these virtualized experiences, where the interaction is less and less a keyboard, and more about speech,” says Frank Chang, a founding partner at AI-focused venture fund Flying Fish in Seattle. The metaverse is also changing the way we interact with technology.

But emerging voice-identity applications in gaming, media, and personal communication are easy to imagine: custom voices for your virtual personas, text messages that read out in your voice, voiceovers by absent ( or deceased ) actors. Today, the most widespread use of TTS is in digital assistants and chatbots. Globally, the text-to-speech market is projected to reach $7 billion in 2028, up from about $2.3 billion in 2020, according to Emergen Research. A search on Google Scholar shows more than 20,000 research articles on text-to-speech synthesis published since 2021. That’s been driving a wave of innovation and investment in AI-powered text-to-speech (TTS) technology. And consumers seem to want tech that sounds more human (although a Google voice assistant that imitated the “ums,” “mm-hmms,” and other tics of human speech was criticized for being too realistic). At the same time, as voice interactions with technology become more common, brands are eager to be represented by unique voices. In one 2020 case, thieves used an artificially generated voice to talk a Hong Kong bank manager into releasing $400,000 in funds before the ruse was discovered. The fear of “deepfake” voices capable of fooling humans, or voice-recognition technology, is not unfounded. And they can easily do the same with pretty much any voice, a prospect that you may find exciting, terrifying, or a combination of both. And the whole thing struck many observers as more than a little creepy-Ars Technica called the demo “morbid.” But Prasad’s revelation of how the “trick” was performed was truly gasp-worthy: Amazon scientists were able to summon Grandma’s voice based on just a one-minute audio sample.

But without knowing Grandma, it was impossible to evaluate the likeness. The woman’s voice that reads a few sentences from the book sounds grandmother-y enough.

Ibm speech to text