Voice Design and Audio Deepfakes - Will You Get Caught Out?

Voice design and voice cloning have BIG implications for identity

Voice Design and Audio Deepfakes - Will You Get Caught Out?

Voice design and voice cloning have BIG implications for identity

Vocal Sovereignty

After an AI fake Drake and The Weeknd track went viral. Grimes said she would split royalties 50/50 on any successful AI generated songs that use her voice. "Feel free to use my voice without penalty", she tweeted, suggesting she was interested in killing copyrights.

Holly Herndon is another artist experimenting with voice tech. In collaboration with Never Before Sounds, she says she makes artwork with her voice. Her doctorate focused on the interplay between machine learning and voice and the implications of the technology for IP and vocal sovereignty. The voice tool that she uses allows anyone to upload polyphonic audio and receive a download of that music sung back in Holly's voice.

Her website says, "you are free to use the material generated as you see fit."

Holly+ πŸ‘­ πŸ—£οΈ 🦾
I’m excited to finally share something I have been working on for the last yearπŸͺ„ Holly+
"The Holly+ model creates a virtuous cycle. I release tools to allow for the creative usage of my likeness, the best artworks and license opportunities are approved by DAO members, profit from those works will be shared amongst artists using the tools, DAO members, and a treasury to fund further development of the tools". - Holly Herndon, Holly+

Voice Scamming

The Guardian published a piece on scammers using voice cloning, suggesting that last year Β£11 million was stolen from unsuspecting consumers who hadn't realised they weren't really talking to their loved ones, doctors, lawyers, etc. but were in fact talking with scammers.

There was a great piece in Bloomberg last month profiling voiceover artists who had accidentally discovered that text to speech services online, were offering their own voices. One company, Revoicer, told Bloomberg News, it couldn't share where it got its voice data from, but that the process was entirely legal. Doesn't sound very transparent to me.

Audio scams are becoming more widespread as cloning tech becomes cheaper. There is a podcast episode of The Future of You about Generative AI coming soon.

Here's an excerpt:

"...audio is improving rapidly and globally we've seen a lot of misuse of audio. It also lacks all the semantic clues that a video has, things you would look at to see if it's wrong.... there's just no surrounding clues, particularly for example, when shared in a WhatsApp group, as in many parts of the world voice messages are how people communicate". - Sam Gregory, Witness

Voice Cloning

ElevenLabs is the most interesting of companies in this field to me.

They claim to have the most versatile and realistic AI speech software, creating lifelike voices for creators and publishers, who are seeking new storytelling tools. Their voice design, as they call it, is offering a text to speech demo on the website.

It's worth going to the website and having a play around with it. Because it serves you have a catalogue of different synthetic voice options. You can choose from accents, nationality, gender, etc.

ElevenLabs || Prime Voice AI

Dubbing tools even let you speak a language that you don't!

In what might be bad news for voiceover artists who do re-voicing, the AI does this automatically, while preserving the actor's original voices across languages.

Speech to speech translation that preserves speaker identity between languages. That could be huge in my view, and it opens up a whole new world of content to people who couldn't immediately make themselves understood further afield

Voice design and voice cloning have BIG implications for identity:

Can others re-present us without our permission?

Can others use our voices to train AI without compensating us?

Is this the death of many traditional advertising practices?

Which me is real and which me is fake?

For more on voice and other news about digital identity listen to the latest episode

More posts from this author