Advertisements
Technology

Project Gucciberg offers classic audiobooks read by an AI deepfake of Gucci Mane

Ever needed to have Leo Tolstoy’s Anna Karenina or Franz Kafka’s Metamorphosis read to you by lure god Gucci Mane, creator of such hits as “Lemonade” and “Wasted”? Effectively, a) that’s an awfully particular need, and b) it’s your fortunate day.

Project Gucciberg is the newest drop from viral manufacturing unit MSCHF, and it does precisely that. Utilizing machine studying, MSCHF created an audio deepfake of Gucci Mane studying a range of classic texts from Little Girls to Beowulf. They’re all free to hearken to and include guide covers that mix in completely with the art work of Gucci Mane’s prolific discography.

The what of Project Gucciberg is luridly simple, however the why is more durable to reply. Should you’re not acquainted with MSCHF, I like to recommend our profile of the outfit from final 12 months. Basically, they’re a bunch of VC-funded creators who make bizarre issues designed to go viral on-line, like squeaky hen bongs and Air Max 97 sneakers full of water from the River Jordan, some of that are bought for a nominal charge. Then they ??? and revenue (presumably by promoting their companies to firms who need issues they made to go viral on-line).

“Gucci’s pronunciation follows a really specific cadence”

Advertisements

Chatting with The Verge, MSCHF’s Dan Greenberg didn’t go into the motivation behind Project Gucciberg however was very happy to speak concerning the mechanics. Audio deepfakes are actually fairly widespread (hearken to this clone of Joe Rogan for a very good instance), to the purpose the place they’ve been used to commit fraud. To make one, you simply want lots of pattern knowledge of your goal talking and the appropriate neural networks to study and replica their mannerisms.

Greenberg says MSCHF collected round six hours of audio of Gucci Mane speaking from podcasts, interviews, and the like. They then created transcriptions of the clips to assist with the text-to-speech (TTS) course of. This required making a “Gucci pronunciation key/dictionary to higher seize the idiosyncrasies of Gucci Mane’s specific argot.”

The redesigned guide covers of Project Gucciberg are a delight to behold.

“Gucci’s pronunciation follows a really specific cadence — he makes use of a a lot better selection of vowel sounds, as an example, than your common TTS reader would,” says Greenberg. “The dictionary breaks phrases up into phonemes (discrete vocal gestures) that our mannequin then makes use of as constructing blocks … So for a easy instance, we want our mannequin to know what syllables to elide, or circulate into one another throughout phrases: it must know to say “talm ‘bout” not ‘speaking about,’ and the Gucci dictionary { T AH1 L M B AW1 T } will get us there the place the written phrases ‘speaking about’ don’t.”

The outcomes are spectacular: the deepfake actually seems like the person himself, although the outcomes usually are not at all times completely coherent or of the best high quality. “Our pretend Gucci Mane usually seems like he’s talking by a foul mic, or over a low-quality web stream, and half of it is because within the coaching knowledge he usually is doing precisely that,” says Greenberg.

Precisely why Gucci was chosen for this mission got here down to 2 elements, says Greenberg: one, the rapper has a particular voice, and two, the Project Gucciberg pun was too delicious to disregard.

The regulation has but to supply an opinion on the legality of creating audio deepfakes of Gucci Mane

Greenberg provides that MSCHF didn’t strategy Gucci to ask for permission to make use of his voice. As a disclaimer on the location slyly factors out, the entire mission raises fascinating questions on copyright within the age of AI fakes. ”We didn’t write the books, and we deepfaked the voice,” it says. “Is that this copyright infringement? Is it id theft? All of the coaching knowledge (recordings) used to make Project Gucciberg have been publicly out there on the internet. Gucciberg lives in that pretty gray space the place every little thing’s new and something goes.” It actually is! The Verge has tried to achieve out to Gucci Mane by way of his report label for a response, and we’ll replace this story if we hear again.

Is Project Gucciberg something greater than a fast click on and a lol? Effectively, not likely. However that’s MSCHF’s enterprise, and so they’re excellent at it. Whereas listening to quite a lot of minutes of the ensuing audio is slightly disorientating, Greenberg suggests there could also be distinctive advantages to the approaching world of on-demand deepfake movie star audiobooks.

“Each on occasion … the acute casualness of Gucci Mane’s narration actually does put the textual content in a brand new gentle,” he says, talking about the advantages of listening to the deepfake model of Kafka’s Metamorphosis. “Gregor Samsa actually comes throughout as simply one other man who doesn’t need to get out of mattress, you already know?”

Advertisements
Show More

Related Articles

Leave a Reply

Your email address will not be published.

Back to top button