Synthetic speech can be a fearful object these days when paired with deepfakes and other AI deceptions, but it’s also an indispensable tool for anyone who can no longer speak on their own. The Acapela Group is keeping a close eye on these people with their new product “My Own Voice” Service that leaves everyone train an AI speech profile for free.
Acapela has been in the text-to-speech space for around 25 years and was recently acquired by tech accessibility giant Tobii Dynavox, although they still operate independently.
Like many industries, accessibility has been heavily impacted by the advent of consumer-scale machine learning. 7 or 8 years ago, Acapela co-founder Remy Cadic recalled, not only was it tedious to customize a synthetic voice for yourself, but the results weren’t great either.
“It was very time consuming – the patient had to exercise for 8 hours. Now we can save a voice with only 50 recorded sentences; It takes about 10 minutes and the voice is ready the next day,” he said. “There is definitely a revolution taking place in neural text-to-speech techniques.”
They weren’t kidding about how quick and easy it is: I went through the new My Own Voice process myself, and it was really just 50 short sentences drawn from a (random, it seemed) corpus of novels and recipe books was drawn , and article. The recording interface was simple and easy to navigate, and in fact my voice was up and running a day or so later. The quality is okay – not scary as some models out there can be, but clearly my own voice (as advertised) and able to handle any phrase I threw at it on the demo site.
Now that it’s here, if I ever need it, I can download it for a fee to use on any compatible speech generation system. This includes, of course, TD Talk and devices from Tobii Dynavox; the company just released a new one last weekin fact – these things are going to be pretty slick.
Delaina tests the new Tobii Dynavox device.
And that’s the real point of all this – it’s not a technical demonstration of the power of neural speech technology, or a demo allowing anyone to feed them a celebrity voice for cloning. It’s a tool specially designed for people who until recently had no options or at best a difficult, complex process if they wanted to keep their voice.
Many who are dealing with degenerative diseases, cancer or certain surgeries know that within a few months or years they may not be able to speak well or at all. Making the process of banking their voice as easy as possible is a service that many will appreciate.
“A big advantage is that we are also adapting the recording for children – we made the recording script easier to read and tuned the system to improve the quality of the synthetic children’s voices. We were the first in the world to do that and we’re still going in that direction,” said Cadic.
Being able to record and re-record or artificially age the bank voice is a new and challenging skill, but one that seems to be yielding results:
Compatibility with offline devices that don’t have the latest neural processing chip is also a key differentiator. “There are online solutions where it’s easy to create a voice, but it’s only available through the cloud and that’s just not practical,” he said.
The company has also found that diversity and thoughtfulness are just as important in the training process as they are in other AI applications. One problem Cadic has pointed out with some super-fast training techniques is that “it’s pretty much just trying to find the speaker in the training material that’s closest to the user. But if there’s no speaker in training that’s close to the original voice, it just won’t sound like it.”
Acapela product manager Nicolas Mazars added that like many AI problems rooted in insufficient training data, this one is not evenly distributed: “This process works well for the average 50-year-old white man, but not for an African American, does it You don’t speak English well. We work in 23 languages and have many users with disabilities. We try to take user feedback and build something for them from them.”
The registration and banking process is free; You can sign up for an account here and train your own synthetic voice in minutes. You only pay if you want to download and install it on a device.
