You’ve probably heard about deepfaking images and videos. That eerily realistic video made with AI? Well, it looks like Meta (previously known as Facebook) has developed a new AI model called Voicebox that specializes in audio. It’s like a powerful text-to-speech system that can create synthesized speech from just text prompts.
What is Voicebox?
At its core, Voicebox is an AI model that creates synthesized speech based on simple text prompts. In other words, give it text and it will read it out in a human-like voice. It’s similar to the text-to-speech functionality you use on your phone or computer, but taken to a whole new level.
One of Voicebox’s strengths is its ability to replicate specific voice styles based on very short audio samples. This is only a 2 second story. This means that you might be able to create synthetic voices that sound like your favorite celebrities or your own voice. It’s like having an on-demand voice actor who will immediately speak whatever you want in your preferred voice style.
Competing AI voice models
give a speech
Speechify and Celebrities are also text-to-speech game players. Speechify is an app that converts any text to speech. You can read books, articles, notes, emails, PDFs, images, and web pages. Speechify also claims to offer audio cloning, audio editing, and audio sampling features. Speechify offers hundreds of free and timeless audiobooks, has a desktop app, and is designed to help people with reading disabilities.
cell phone meta logo (Costfoto/NurPhoto via Getty Images)
Mark Zuckerberg ‘Twitter killer’ thread infuriates users with massive data collection: ‘Privacy is almost zero’
Eleven Lab
Meanwhile, Eleven Labs is a startup that uses AI to generate synthesized speech with contextually relevant emotion and natural language understanding. It provides a platform for creating and customizing high-quality voices in any voice and style for various industries such as video games, animation, digital assistants, education, entertainment, advertising, and podcasting. It also has tools to detect synthesized speech and verify its authenticity. Eleven Labs works with actors who provide voice samples and gets paid when their voice clones are used. They use their own deep learning models to create AI-powered speech.
Both are very good, but they are not as versatile as Voicebox, which can imitate a real voice from just a few seconds of audio. It’s like comparing a Swiss Army knife to some really good spoons. Each has its uses, but one is more versatile.
the power of voicebox
But it does more than just create falsettos. Voicebox can also clean up audio by removing annoying background noises (such as a barking dog while recording). And it’s not just about English. The AI also speaks French, Spanish, German, Polish, and Portuguese, and can also translate sentences from one language to another while maintaining the same voice style.
Come on, SIRI: Apple’s new audiobook AI voice sounds like a human
The Meta (formerly Facebook) logo marks the entrance to its headquarters in Menlo Park, California on November 9, 2022. – Facebook owner Meta plans to lay off more than 11,000 staff in ‘most difficult change in Meta’s reform’. It’s history,” boss Mark Zuckerberg said Wednesday. (Josh Edelson/AFP via Getty Images)
Meta’s Voicebox: Breakthrough or Threat?
Unfortunately or fortunately, depending on their position on AI, Meta has no plans to open source Voicebox anytime soon. That’s why people wonder if I’m trying to avoid potential problems. For example, AI voice technology can be used negatively, such as in harassment campaigns. Or maybe Meta has future plans to make money off of this model.
Source of Voicebox’s vast training data
One of the interesting things about Voicebox is that it was trained on a massive data set of over 60,000 hours of audio from English audiobooks and another 50,000 hours from multilingual audiobooks. According to Meta, it used public domain audiobooks as its primary data source, but also used other sources such as podcasts, speeches, and radio programs. However, using public domain audiobooks comes with some challenges and limitations, such as quality, consistency, coordination, and speaker identity. Meta claims to have addressed some of these issues in data processing and model design.
more for me security warning, Visit and subscribe to our free Cyberguy Reports newsletter. CYBERGUY.COM/NEWSLETTER
Double-edged sword of technology
President Obama reverses ‘stupid’ court order after judge blocks communications between Biden administration and social media firms
The rise of AI voices is a bit of a thorny subject, especially for voice actors and, more recently, writers. They are concerned about companies using AI to synthesize speech without paying for it. With the audiobook market growing significantly and companies constantly looking to cut costs, this could pose a new problem for audio professionals.
But don’t get me wrong. It’s not just about work. There are some real concerns about the extent to which fake voices are used for fraud. For example, there was an incident in which a synthetic voice impersonating a CEO was used in a large-scale robbery. There are also concerns that deepfake voices could be used to sabotage voice biometric systems, such as those used for online banking.
This technology sounds cool, but it also has a dark side. Imagine getting a call from your boss asking you to transfer a large amount of money to close an account. As a boss, I do what I’m told. However, it wasn’t. That’s right; it was an AI-generated fake synthetic voice that looked exactly like the boss. Wild, isn’t it? But this is not the plot of the movie. It actually happened! This was his one of the first cases where a fake voice was used in a robbery, and it left law enforcement and his AI experts scratching their heads.
Kondo was optimistic about the future of artificial intelligence. (Jakub Porzycki/NurPhoto via Getty Images)
Dal-2 VS. BING Creator – Who comes out on top in this AI showdown?
And it’s not just robberies. Deepfake voices can be used to trick systems that rely on speech recognition. We’re talking something like online banking that uses the user’s voice as a form of identification. If criminals can create a convincing fake voice of you, they may gain access to your account. It’s a bit like forging a signature, but uses your voice instead.
Combat the threat of deepfakes
So while we marvel at what technology can do, it’s also important to be aware of potential risks and stay one step ahead. It’s a high-tech cat-and-mouse game, with AI experts and companies working hard to spot and stop these deepfake voices before they can do any harm.
Fortunately, there are people trying to fight back against the potential abuse of deepfake audio. For example, some countries have started enacting laws to regulate deepfakes. There are also projects like the Automatic Speaker Verification Spoofing and Countermeasures Challenge (ASVspoof), where scientists and engineers are working on ways to counter deepfake voice attacks.
Cart key points
We live in an age where technology is evolving at breakneck speed, changing the way we work, communicate, and even listen. The possibilities of AI like Meta’s Voicebox are undoubtedly exciting, but they also clearly need to be approached with caution. There is a fine line between innovation and aggression, a balance we have all yet to find.
Experts argue that the difference between Chinese and U.S. AI investments lies in the fact that the U.S. model is driven by private companies, while China takes a governmental approach. (Josep Lago/AFP via Getty Images)
CLICK HERE TO GET THE FOX NEWS APP
With all these advances and potential risks, what do you think of the future of AI and deepfake technology? Do you see it as a boon or a calamity? Cyberguy.com/contact
For more information on security alerts, please subscribe to our free CyberGuy Reports newsletter at the link below. Cyberguy.com/Newsletter
Copyright 2023 CyberGuy.com. all rights reserved.