AI-Powered Voice Conversion: The Latest Controversy in Music

The music industry has seen a lot of technological advances in recent years, with the newest addition being the use of AI. One particularly fascinating application of this technology is the AI ​​Diff SVC (Artificial Intelligence Differential Source-Channel Voice Conversion), which has been used to create impressive covers of popular songs by Latino artists such as Bad Bunny and Duki.

The AI ​​Diff SVC is a voice conversion system that can change the vocal characteristics of a recording while keeping the original pitch and melody intact. In essence, it can make one singer sound like another, or even make a singer sound like a different version of themselves. This technology is based on deep learning algorithms, which can analyze the unique characteristics of a singer’s voice and then use that information to manipulate other recordings to match their vocal style.

In recent weeks, many videos created by social media users have gone viral on the internet, in which songs are changed to make it seem as if other artists are singing them. This has sparked the curiosity of many fans, causing them to wonder what their favorite singers would sound like if they did covers of other songs, or at least collaborated with the song’s original artist. However, others are worried that AI will eventually take over not only music but many other aspects of the industry and life itself.

In the case of Bad Bunny, the AI ​​Diff SVC was used to create a cover of “La Curiosidad” by Puerto Rican singer Jay Wheeler. Using this technology, Bad Bunny’s voice was transformed to sound like Jay’s, resulting in a truly unique and captivating rendition of the song. The cover quickly went viral, and it’s not hard to see why – it’s an exciting example of how AI can be used to create new and exciting music.

Bad Bunny’s voice was also used in the song “Saoko,” by Spanish star Rosalía, where not only the vocals were altered but also the whole gender. The result quickly went viral as fans loved the rendition.

But swapping genders isn’t the only change AI is making. Several videos have gone viral of English-speaking singers such as Drake singing some of Benito’s songs, while the Canadian rapper’s voice was used in the song “Amorfoda,” in a unique Spanglish rendition. Meanwhile, Ariana Grande’s feminine, high-pitch voice was used in the song “Moscow Mule.”

https://www.youtube.com/watch?v=1gziSNUc_OE

Similarly, the AI ​​Diff SVC was used to create covers of classic Latin urban music songs with Argentinian rapper Duki’s voice, who has a very unique pitch. Now, we have new versions of “La Curiosidad” by Jay Wheeler and Myke Towers, “Amorfoda” by Bad Bunny, “Friki” by Feid and Karol G, and many more, thanks to AI. 12 renditions have been published on the @sukrynt TikTok account and have also gone viral on other social media platforms, where they accumulated millions of views.

Since late last year, a plethora of songs sung by English-speaking artists modified by AI has surfaced online, like Ava Max singing “Flowers” by Miley Cyrus or Ariana Grande revitalizing Dua Lipa’s “Levitating.” However, Diff SVC has also been used to “bring back” dead artists, such as the video of Michael Jackson singing Daft Punk’s “Get Lucky,” which quickly went viral.

Through this program, fans have found millions of possibilities and alternatives to create dream versions of their favorite songs by any singer. But it has also brought the topics of ethics and intellectual property into debate, as well as social media bans, as many users had their profiles suspended after publishing these versions, deemed “illegal” in the eyes of the record companies and the artists they represent.

On January 22nd, the developers of DiffSVC announced that they would withdraw the software to prevent abuse and legal problems. But the program has been copied and is easily located on the web, and some users have started to make their own versions of songs. Even though this is also illegal, it is not being penalized on the internet.

Artificial Intelligence is not perfect and for now, it is not capable of accurately recreating the quality and emotion of a human voice. But it’s still a fun experience if used in moderation, equally promising as it shows how far technology is advancing today.