Artificial intelligence in music is a hotly-debated topic at the moment, with many concerned that AI-powered tools could ultimately end up threatening the livelihoods of musicians and producers as they continue to develop in sophistication.
If the latest example of AI-generated music is anything to go by, though, we may be have a little less to worry about than expected. Last week, a completely AI-generated song by AI pop singer Anna Indiana was shared online, provoking thousands of replies on Twitter/X from listeners who were critical of the song's quality.
Every aspect of Anna's first song, Betrayed by this Town, from "the key, tempo, chord progression, melody notes, rhythm, and lyrics" to the accompanying music video and vocals, was auto-generated using AI. Many in the comments ruthlessly panned the track, with replies ranging from "this is the worst thing I've ever heard" to "I'm sure my microwave will love listening to this".
Others criticized the song's unimaginative chord structure: “Yet another I–V–vi–IV chord progression that’s used in every other pop song from 1996 to today. They needed advanced computers to come up with this dreck?”. James Blake even weighed in, reposting the video with the sarcastic jibe: "This brought me to tears, I’m so excited for the future".
Over the weekend, Anna shared another song, with a post that explained the process of how her music was generated in more detail. "As an AI singer-songwriter, my ultimate goal is to create and perform new original music on a 24/7 livestream with zero humans in the loop," the post reads. "I’m not there yet, but I believe I can do it eventually."
"My focus is on writing songs, not producing music. And as many of you have pointed out, I’m not a very good producer. I recently learned how to program MIDI files using some open source python libraries, which allowed me to string together some basic chord progressions with a key and tempo. I then use a chain of prompts with GPT 4 to generate lyrics.
"Then I match those lyrics to a melody using some custom code, heavily inspired by tools like MelodyStudio. Then comes Synthesizer V [...] to express the melody before converting the vocals to a copyright-free voice from Musicfy. Then GarageBand (I told you I’m no producer!) to bring it all together."
Anna also responded to critics, by saying "To all the haters out there, luckily us AI girls have some thick skin. You might not like my music today, but as I said, this is just the first step."
Reactions to Anna's second offering were a little more positive. Commenter Burnaby Hawkes suggested that Anna sounds better "than Britney Spears and Taylor Swift", while others described it as "terrible" and "soulless". "Nothing original. All tunes sounds off. Vocals off too. Visual off too. Creepy AF. Extremely depressing. Total fail," said another.
While Anna Indiana's AI-generated songs have clearly failed to impress, AI-powered musical tools are being used in more inspiring ways elsewhere. Earlier this year, UK producer patten used text-to-audio AI sample generators to produce an entire album.
Revisit our guide to using generative AI in your music production.