They use the ElevenLab tool to imitate the voice of celebrities and generate audio with racist speech

MADRID, 31 Jan. (Portaltic/EP) –

The technological ‘startup’ ElevenLab has warned of the bad uses that you have detected from your voice clone toolafter detecting the impersonation of celebrity voices, which have led the technological ‘startup’ to introduce additional security mechanisms.

During the weekend a series of audio clips apparently starring celebrities like Joe Rogan, Ben Sharpio and Emma Watson, with racist and violent content.

According to the investigation carried out by Motherboard, it is a series of audios uploaded to the 4Chan platform, which have been created with the artificial intelligence tool from ElevenLab so that the voices sound like those of the artists.

This misuse has been recognized ElevenLab in a post on Twitter. “While we see our technology being overwhelmingly applied to positive use, we also see a growing number of cases of misuse of voice cloning“.

From the ‘startup’ they assure that can track any audio generated with your tools up to the creator as all clips are watermarked. But they have also announced that they will implement a series of additional measures with which they hope to tackle the problem.

At the moment it is about ideas, but they consider requesting an additional account verification to use your voice clone tool, such as payment details or an official identity document. Too request a sample with text to verify voice copyright, and remove Voice Lab entirely and manually verify each clone request.

The technology company has two tools that allow voice cloning: Instant Voice Cloning, within Voice Lab, which only needs a small sample but offers less quality; and Professional Voice Cloning, which requires more examples of a specific person to be able to faithfully reproduce their voice thanks to a trained model.

The appearance of audio clips that impersonate the voices of celebrities delves into the potential misuse that can be made of this type of tools based on artificial intelligence. Recently, the OpenAI language model, ChatGPT, has stood out for being very efficient when it comes to generate coherent texts with a high writing capacitynot allowing in many cases to distinguish if a person has written it or if this tool has been used.

Previously, the applications of ‘deep fake’, of manipulation of audiovisual contenthave generated alarm due to the degree of realism with which artificial intelligence was capable of impersonate a person and to deceive those who watched the videos.

Related articles

Gam Gam Ganesha: 25 Tamil directors approached me: Anand Deverakonda

Anand Deverakonda impressed the audience as an actor with the youthful entertainer 'Baby'. Currently he is getting ready for the release of his...

ANSES pays the Unemployment Benefit: who receives today, Friday, May 24, 2024

The National Social Security Administration (ANSES) begins the payments corresponding to the Month of May for unemployment insurance holders. Today, Friday the 24th,...