The other day, I came across a short video with a song playing in the background. It was a cover of a song, by JuiceWRLD, a deceased rapper. I enjoyed the small portion of the song that I had listened to, so, curious to hear more of the song, I searched for the full version. To my surprise, the person singing wasn’t JuiceWRLD. In fact, the person singing wasn’t a person at all, it was an AI-generated voice made to sound like him, and it had fooled me.
The process of voice-cloning begins with a voice sample. The sample can be as short as 3.7 seconds of speech for Baidu’s Deep Voice software to effectively synthesize the voice, but larger and higher quality samples promise better results. Afterward, the software analyzes every nuance of the voice and creates a dataset on which to train a new voice model. Once complete, convincing fake audio clips can be made with the model. The uses of such technology, both ethical and unethical, are extensive. However, one of the most interesting applications of voice cloning is in music. Voice-cloning technology allows fans to hear their favorite artist, whether they’re dead or alive, on any song.
The potential of AI in the music industry goes beyond voice cloning, though. By analyzing a large dataset of music AI can even create new compositions that artists can use directly or use as inspiration. In the same sense, AI tools like ChatGPT can write original lyrics with a given topic and genre. Providing artists with inspiration in writing their own lyrics.
AI’s capabilities raise the question of whether or not it will completely replace producers’ or artists’ roles in the production of music. I believe the answer is no due to the nature of AI. AI revolves around imitating humans by analyzing large datasets. As a result of this, AI is unable to create anything that is truly original, it can only mimic human creativity. Regardless, it is undeniable that AI is a tool that has the potential to revolutionize the music industry.
See more like this:
