How Facebook Is Using AI to Improve Voice Synthesis
In this article, take a look at how Facebook is using AI to improve voice synthesis.
Join the DZone community and get the full member experience.Join For Free
At this point, you know that Facebook has proven time and time again to be one of the most innovative companies in recent memory. Their team took a relatively simple product and turned it into a world-renown brand with over 2.45 billion active users.
Recently, the social media team announced that they had improved their text-to-speech synthesis technology. This news is significant for those who use AI to run their businesses and create new products.
We will take a look at several practical uses of text-to-speech technology, then explain the new update from Facebook, and what this means for the tech community.
Case Uses for Voice Synthesis
Essentially, text-to-speech synthesis is an accessibility feature that can improve the lives of people around the world. Before the explosion of AI software, TTS was known as concatenative to speech.
Concatenative-to-speech required people to manually load in speech fragments from the same speaker. When someone typed an input in the program, the pieces get placed together to form words. If you ever used one of these programs, you know how unnatural and garbled the text can sound, depending on complexity.
TTS synthesis opens plenty of new doors for business owners and alike. The most sought after use is as a simple reading application. Imagine finding an article online, clicking a button, and being able to list to the text within the content while you’re working on other tasks. The possibilities for consumer marketing and engagement would open up like never before.
This type of speech is also great for educational purposes. Subscription-based learning management systems are widespread across all industries. It’s estimated that the LMS industry value will hit 22.4 billion by 2023. Speech software allows LMS developers and site owners to open their site up to the visually impaired, and gives users a new way to experience their service.
Facebook’s TTS Advancements
Facebook announced that they created a new type of text-to-speech AI software that will make waves across multiple industries. The new system uses data collection software to quickly and efficiently translate text to speech in a variety of styles and languages.
At this point, the program works in English, French, German, Italian, and Spanish. The unique styles, including fast, formal, and soft, will allow the user to customize the way the speech is relayed to them, which adds even more accessibility to an already useful feature. It’s also worth noting that the program can also create and use various accents.
One of the most notable features of this new system is translation speed. Previously, systems would need hours of data just to form a couple of seconds of speech. The training process was rigorous, and a simple mistake could be disastrous.
Using the new software, Facebook could generate 1 second of text in an astounding 500 milliseconds, making this system the most efficient TTS software to date. It’s worth noting that the time to train is dramatically shorted due to this advancement.
Facebook’s core system uses four different elements to make the speech from text work. Each system is layered and connected to other sections to create a web that funnels and organizes text in ways that will fundamentally shape TTS forever.
There’s no doubt going to be plenty of exciting advances in this software by the time it becomes widely available. Facebook announced that they have plans to add more languages and accents, but that’s just the tip of the iceberg.
Once this program is in the hands of more users, Facebook will be able to make improvements and changes as time goes on. Machine learning focuses on data, and the more people using your software, the more data you have for future testing.
We can expect to see this software translate exceptionally well to mobile users. Facebook consistently innovates when creating experiences for mobile users. There’s a good chance that this is because 96 percent of users across Facebook use the platform on their mobile devices.
It’s exciting to see that the future of AI is looking bright. Expect to hear more about this update as Facebook gathers more data internally.
Opinions expressed by DZone contributors are their own.