Veritone launches new platform to let celebrities and influencers clone their voice with AI

Recording advertising and product warranties can be beneficial for celebrities and influencers. But is it like too hard work? That’s what the American company Veritone bets on. Today, the company is launching a new platform called Marvel.AI that allows creators, media figures, and others to deep fake their own voices to license them as they wish.

“People want this deal, but they don’t have enough time to go to the studio and create the content,” says Ryan Steelberg, President of Veritone. The Verge. “Digital influencers, athletes, celebrities and actors: this is a huge asset that is part of the brand.”

With Marvel.AI, anyone can photo-realistically copy their voices and distribute them as they see fit. While celebrity Y sleeps, their voices go out, allowing them to record radio broadcasts, read audiobooks, and more. Stillberg says the platform could also use archival recordings to train AI models to revive the voices of the dead.

“Whoever has the copyright on that voice will work with them to bring it to the market,” he says. “It will depend on the rights holders and what they think is appropriate. But hypothetically, yes. You can get Walter Cronkite to read the nightly news again.”

Speech synthesis has improved rapidly in recent years, and machine learning techniques can generate more realistic speech. (Think of the difference between what Apple’s Siri sounded when it was released in 2011 and what it is now.) Many big companies like Amazon offer off-the-shelf text-to-speech models that generate robotic yet massive speech. Not offended. However, the new company is making boutique voice replicas that sound like specific individuals, and the results are almost indistinguishable from the real thing. For example, listen to podcaster Joe Rogan’s voice duplication.

What motivated Veritone to create Marvel.AI is not only this leap in quality, but also the potential of synthetic speech to combine with the company’s existing business, says Steelberg.

Veritone markets itself as an “AI company,” but a large portion of its revenue appears to come from outdated advertising and content licenses. As Steelberg explains, the advertising subsidiary Veritone One is investing heavily in the podcast space and deploys more than 75,000 “ad consolidations” each month with influencers. “It’s mostly a basic integration like PPL,” he says. “I have a talent for sponsorship and advertising. Very effective, but costly and time consuming. ”

Another part of the company, Veritone Licensing, licenses videos from several major archives. This includes archives owned by broadcasters such as CBS and CNN and sports organizations such as NCAA and US Open. “When you see an Apollo moon landing scene in a movie or the word Tiger in a Nike commercial, everything goes through Veritone,” says Steelberg. It is the licensing and advertising experience that gives Veritone an edge over purely technology-focused AI startups.

For customers, Marvel.AI offers two streams. One is a self-service model where anyone can choose from a catalog of pre-generated voices and generate voices on demand. (This is how Amazon, Microsoft, etc. have been doing this for years.) But the other trend, “focus,” says Steelberg. There will be a “managed white glove approach” for customers to submit training data. Veritone creates negative clones for these. The resulting model is stored on Veritone’s system and can generate audio when the customer wants it. Marvel.AI also serves as a marketplace where potential buyers can submit requests to use these voices. (It’s not yet clear how all of these prices will be priced.)

Steelberg is confident that the demand for this voice exists and that Veritone’s business model is ready. However, the success of Marvel.AI will determine one major factor: the quality of the AI ​​voice that the platform can produce. And this is much less certain.

When asked for an example of the company’s work, Veritone took three short clips. The Verge, Each single line warranty for the Mint brand. The first line reads Steelberg himself, the second line reads an AI clone, and the third line reads a gender-changed voice. You can hear all three below.

The AI ​​clone is a pretty good counterfeit, at least for my ears, but it’s not a perfect copy. It is flatter and cut off than it really is. But it’s also not a complete demonstration of what the voice can do during the warranty period. Steelberg’s delivery lacks the enthusiasm and sincerity you can expect for real advertising (we’re not doing him wrong, he’s an executive, not a voice actor). Thus, Veritone’s voice model appraised the full range.

It’s also not a good sign that the voiceover of the platform’s sizzling reels (included at the top of the story) was done by Steelberg himself, not an AI copy. The company thought that voice replication wasn’t enough for the job, or it ran out of time to create the replication. Either way, it’s not a great support for the product.

But the technology is evolving rapidly, and Steelberg highlights that Veritone has the resources and expertise to adopt the new machine learning models to come. What differentiates itself, he says, is that it manages the customer and client experiences to actually deploy large-scale synthetic voices.

One problem Steelberg offers is how synthetic speech dilutes the power of warranties. After all, the attractiveness of a product warranty depends on the belief (but delusional) that this or that celebrity really enjoys this particular brand of cereal/toothpaste/life insurance. If celebrities can’t express their support for themselves, won’t they take away the sales power of advertising?

Steelberg’s solution is to create an industry standard for disclosure. That is, it is a kind of audible tone that is played before a synthetic speech, a) tells the listener that it is not a real voice, and b) reassures that the owner of the voice approves this use. “Not only avoiding negative meanings that deceive consumers, [this or that celebrity] I really approve of this composite content,” he says.

It is a question like this that will become more and more important as synthetic content becomes more common, and it is clear that Veritone is thinking hard on these issues. Now the company just needs to persuade influencers, athletes, actors, podcasters and celebrities to make their own voices.


Please enter your comment!
Please enter your name here




India’s Top 10 Leather Shoes Brands

Are you looking for the most popular...

US sanctions a Chinese facial recognition company with Silicon Valley funding

The U.S. Department of Commerce has sanctioned 14 Chinese tech companies for human rights violations against Uyghur Muslims in Xinjiang, one...

How Enterprise Mobile App Development is Transforming Your Business –

You may have heard of the term enterprise mobile app development for the first time. It is widely used in the context...

MSI Summit B15 review: average business

This year, several companies, primarily known for their gaming laptops, have entered the portable business and productivity segments. It's already...

Microsoft is finally fixing Windows 10 rearranging apps on multiple monitors

If you're a Windows user and have multiple monitors, you've probably experienced the headache of randomly rearranging apps when you resume...