ElevenLabs adds AI voice of celebs to new digital narrator — but is it safe?

8 Min Read

Don’t miss OpenAI, Chevron, Nvidia, Kaiser Permanente, and Capital One leaders solely at VentureBeat Rework 2024. Achieve important insights about GenAI and broaden your community at this unique three day occasion. Study Extra


Every week in the past, ElevenLabs, the AI voice startup based by former Google and Palantir engineers, made headlines with its first main consumer-centric product – a Reader app.

At present available on iOS, the product is a devoted voiceover resolution that converts any textual content file or hyperlink from the online into AI audio, narrated in several AI voices and accents. Right now, the corporate introduced it’s increasing this library of voices on the app to incorporate AI voices of late Hollywood celebs Judy Garland, James Dean, Burt Reynolds and Sir Laurence Olivier.

The corporate has partnered with CMG Worldwide, the agency managing and defending the mental property rights of dwelling and deceased celebrities, to recreate and launch the enduring voices. Moreover, it plans to construct on this work with many extra celebrated AI voices set to launch within the coming months. 

Reader provides AI voice to any digital textual content

Whereas ElevenLabs has particularly centered on the inventive business with AI fashions for text-to-speech and speech-to-speech conversion, dubbing and sound impact creation, the Reader app provides a extra tailor-made kind to its analysis within the text-to-speech area. All a person has to do is give the hyperlink or file for any digital textual content – be it an article, PDF, publication or 300-page e-book – and the app immediately processes the textual content and begins the voiceover AI narration, with a inexperienced highlighter following alongside and highlighting every phrase spoken by the AI.

See also  Top 7 Realistic Voice Generators for Stellar Audio Content

The characteristic is obtainable in English, though customers can customise their expertise by selecting from 11 voices and accents, from male to feminine, American to Austrian to British English. Now, the Iconic voices launched right now provides to this expertise, permitting customers to find and expertise content material within the voice of the late stars. 

Think about a person having the ability to take heed to L. Frank Baum’s The Great Wizard of Oz within the voice of late Judy Garland who acted within the cinematic adaption of the novel.

For the relations of the late stars, the AI-based voice recreation is a chance to ensure that the celebs’ legacies reside on, with their current followers getting a option to reconnect with them, and new-age customers getting a option to uncover them. In the meantime, for ElevenLabs, the announcement is anticipated to drive extra engagement on the brand new app.

“Judy Garland, James Dean, Burt Reynolds and Sir Laurence Olivier are among the most celebrated actors in historical past. We deeply respect their legacy and are honored to have their voices as a part of our platform,” mentioned Dustin Clean, head of partnerships at ElevenLabs “Including them to our rising record of narrators marks a significant step ahead in our mission of constructing content material accessible in any language and voice.” 

See also  IBM, Meta launch AI Alliance with over 50 tech members to advance 'open, safe, responsible' AI

Are these AI voices protected from abuse?

One of many greatest issues related to voice cloning expertise – just like the one at play right here – is that voice recreations of recognized personalities can painting them as saying issues they by no means really mentioned in the true world. Biden’s Robocall incident is the largest instance of such a problem. In the identical manner, what if a CEO’s voice is cloned to make them say issues that would doubtlessly break their or their firm’s status?

ElevenLabs says it understands these issues and is transferring to broaden partnerships for the enduring voices characteristic with a specific give attention to security.

Sam Sklar, who handles development advertising and marketing at ElevenLabs, advised VentureBeat that the corporate retains full management over celeb voices and makes them out there solely on the Reader app, which has been designed in such a manner that customers can solely convert digital textual content into AI narration for particular person consumption — reasonably than additional sharing or downloading.

“For instance, by means of the Reader App, you could possibly select an article on VentureBeat and choose Judy Garland to relate it only for you. You can not entry her voice by means of the ElevenLabs voice library (a separate net product of the corporate). This implies they’ll’t be used along with our typical text-to-speech instruments on the platform, nor can the content material they communicate by means of the Reader App be downloaded or shared,” he defined.

If a person uploads dangerous content material as textual content to document its iconic voice narration by means of a secondary system, the corporate won’t even generate the AI voiceover. It has positioned automated and human moderation processes in between to establish and block hate speech and different types of textual content that violate its phrases of service.

See also  Zephyr: Direct Distillation of LLM Alignment

As for the possibilities of the voice library being misused to clone celeb voices from scratch, Sklar says the platform has been constructed with a number of safeguards, together with a voice captcha verification that matches the audio samples uploaded for cloning with the voice recording of the person. If the voice doesn’t match after a couple of makes an attempt, the cloning request isn’t processed. There’s additionally a “no go” voices coverage in place, which prohibits the cloning of voices deemed excessive danger. 

“Any try to clone these voices might be blocked,” Sklar mentioned.

Whereas these steps do scale back the possibilities of celebs, actors and enterprise executives’ voices being cloned, there nonetheless could be circumstances of violations. For example, malicious customers might craft the content material for the Reader app in such a manner that it bypasses the moderation measures positioned by the corporate. 

In the long term, will probably be attention-grabbing to see how the enduring voices functionality, which has been positioned as an providing for followers and lovers, impacts the business. The Reader app internet hosting will probably be rolling out each globally and to Android units this summer season. Help for extra languages can be on the best way.


Source link

Share This Article
Leave a comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Please enter CoinGecko Free Api Key to get this plugin works.