Imagine how surprised I was when one of my favorite singers dropped a new song after a hiatus of six years. Dean is a Korean R&B singer-songwriter who has not released any solo in a long time. To my disappointment it was a cover song but something was better than nothing at all. However, it was actually an AI generated cover song which surprised me because it bore an uncanny resemblance to his voice. Hence the video was not uploaded by an official account, though it racked up over 2 million views on the platform. Using AI to mimic artists’ voices was at that time not exactly unknown to me, but it usually did not sound 100% real. After a few notes, you could hear it was AI generated which was not in this case. I wondered if the user who used Dean’s voice unauthorized would be liable for copyright infringement. Thus far the video is still up and running. Further research showed that people are not only using AI to make cover songs, but also ‘original’ AI generated songs with voices from actual real artists. One AI generated Drake song actually went viral and had millions of views and streams before being taken down from various platforms.
Popular websites include Voicify.ai which is a website that provides paid services to generate a realistic cover song with your selected vocals. Trending choices range from cartoon characters like Eric from South Park to Michael Jackson. A quick search for free platforms led me to Covers.ai, although the selection of voices is limited compared to paid websites. I could create one cover song with the free trial. Unsurprisingly it did not sound as flawless as the covers you can find on TikTok, but oddly enough it did get better at the latter part of the song.
These ongoing unofficial releases have sparked a discussion about generative AI and its ethical implications. Issues concerning copyrights, originality and integrity are raised. The fact that AI can replicate our voices so easily gives me a sense of foreboding for the future. For artists who depend on their voice for a living, the rise of AI may be threatening to them. Since there are no regulations yet, this kind of AI generator could also be used for negative purposes such as voice phishing.
With all these implications, can we still enjoy generative AI music responsibly? One thing for sure is that we must find the right balance between innovation and integrity. Perhaps generative AI is the way to provide us with music from our favorite inactive artists (any Frank Ocean fans?). Let me know if you have ever been fooled by an AI generated cover song and thought it was real.
I was thrilled and a bit disappointed when I saw your post haha because I was planning to write about AI voice covers too. I use an AI tool quite often to make my favorite singers from my countries cover good songs released in different countries. I also use it to listen to trendy music sung by artists who have passed away, like Whitney Houston.
I think this topic is timely and has been both sensational and controversial in the music industry. It’s understandable that ethical concerns have been raised, especially for artists who have gone through countless practices and training to earn their voices and skills.
Given the circumstances, It seems necessary to consider the responsible coexistence of generative AI with human creativity. Generative AI can contribute to the music industry by adding creativity and diversity in a way of securing creators’ or artists’ rights and other ethical issues.
(Resubmission of comment due to wrong displayed name)
First of all, I did not know this existed, but I must say I have that a lot of the time with AI applications. I wrote my blog about voice cloning, so I could have expected it.
It must feel very threatening for the music industry songs now can be made so easily.
My personal option would be that these AI could be used for personal use, use like Juyeon uses it, but for uploading it. The artists owns their own voice. think it is different when we using a artists voice that isn’t alive anymore, however than it should be limited to family or friends (maybe artists need to include the ownership of their voice in our will in the future which is a weird thing to think about).
Thank you for the interesting post. Personally, I have never been fooled by AI generated cover songs (yet). However, I would feel quite disappointed especially in the case of Dean. In addition, I would feel uncanny when ‘new’ songs of a deceased artist would be posted/shared. However, I think it all comes back to ethics, could you listen to ‘an artist’ that is non existent? Could you listen to a song from your favourite artist, but he/she never sang the song? And indeed what about the copyright? Many questions that are still unanswered. Overall, a very intriguing topic!
Cool post, I never considered the idea of fans generating content of their favorite inactive singers using AI. I think the implications become a bit murky when you have people using it to generate songs of deceased artists, but I do believe that as long as something is not monetized/commercialized, then it should be allowed. For example, holograms: there was a huge controversy when a hologram of Tupac first showed up on a festival stage. Now people don’t really think about it so much anymore when holograms of deceased artists are used. I think it will be much the same with AI – it feels weird at first, but society will normalize it sooner than later.
Very interesting blog post!
I feel that the ability to replicate the voices of our favorite artists is truly a blessing. Think about the opportunities for collaborations and new songs. I am a huge rap fan, and two of my most played artist have passed away. For me, it is really exciting that there is a chance Tupac and Mac Miller will have their verses on the same song. The prospect that artists can continue to produce music after retiring gives the musical world a completely new dimension. Despite ethical concerns, it’s intriguing to hear AI-generated cover songs that sound so authentic.
Thank you for your interesting post! As a huge Frank Ocean fan, I would not be mad for being able to listen to AI-generated songs that replicate Frank’s voice. My take on AI-generated songs and their ethics is that as long as the songs are not published for monetarization, we should be able to enjoy them freely. In the case of publishing the songs for profit, I feel like copyrights would/should be applicable. For me, listening to an AI-generated Frank song would never compare to the “real deal” even after a long hiatus, and never would I pay for it.