AI Audio Adventures (Or How I Asked Holly+ To Eat My Song)

Music by DALL·E

In my inquiry around AI and Audio, I stumbled upon this interesting platform by musician/performer/experimental artist Holly Herndon. Her application — called Holly+ — takes an uploaded audio song and transforms it through AI and Voice into her unique musical style. I had to try it out (of course) and the results are strange but interesting.

I used a demo from a song of mine — with some lyrics from my drummer friend, Bob — called Faucet Drop (Quarantine Together).

The Holly+ tool digested the audio file and turned out a very different remixed version, for sure, but with the underlying DNA of the original demo still intact. There are no recognizable “words” in her vocal AI-infused audio, but that’s fine, as it becomes a different kind of art and collaboration.

 

But I also wanted to take it another step forward, by bringing me back into the mix (so to speak), so I sampled the first section of the Holly+ remix of my song, and began to make another short sample remix, adding some other elements on top, giving it a little more disorientation. In this one, it was all sampling — sections and loops, gathered together and the feel is very different.

So this moved from me to her to us.

Peace (Singing It),
Kevin

PS– Here is Herndon giving a TED Talk about AI, voice and more. (By the way, she is part of a technical team working on ways to protect artists’ intellectual property in the Age of AI through the work of Spawning, and its website: Have I Been Trained? where you can search for art that has been scraped into AI databases.

AI Hip Hop: Do You ETMOOC Too?

ETMOOC AI Rap

Yesterday, I was playing around with the AI Music application within the voice-centered AI site called Uberduck, and I went back in to play around today, generating a teaser verse of a song about ETMOOC2. The AI generated the lyrics from a prompt, and then added a voice, rapping the words over a beat.

Listen to the short one-verse rap

Peace (too),
Kevin

ETMOOC: The Emergence Of Deep Fake Music

AI Music by StableDiffusion

I’ve been paying attention to the ruckus over the deep fake music that has showing up on social media in the last few weeks, where an AI-influenced song – Heart On My Sleeve — using the voices of Drake and The Weeknd has shaken the music industry. The song in question was different from some of the earlier memes and remixes that were taking root in social media. In this case, it was an entirely new song by AI with the voices of the two pop stars singing new words generated by AI. And it was pretty authentic sounding, too, in terms of song theme and vocal intonations and phrasings of the two artists.

This podcast episode — The Daily from New York Times — is a good listen for an overview of the situation.

As a musician and songwriter, I find myself conflicted on the emergence of AI in the field of making music. On one hand, it opens up some interesting doors for creative composition, using sounds and voices and techniques that might not otherwise be easily available to musicians. I’ve done some explorations of AI music sites but haven’t yet been all that impressed by what I’ve found. Clearly, though, there’s more out there that I haven’t yet discovered and played around with.

On the other hand, the legal and ethical issues of copyright and intellectual property use and infringement are huge, as it has been for the AI art generation field (lawsuits are already underway over the scraping of online content to feed the AI databases) and it feels like another reminder that the AI companies will need to find mechanisms (or be forced to, via lawsuits) for recognizing the contributions of musicians to any field of AI Music. I don’t know how that will be done, but it seems important for the tech folks to figure it out.

This all reminds me a bit of an earlier post I wrote a few months back about using loop tracks to construct songs — which is something I do along with traditional guitar-in-hand songwriting– and whether that is “songwriting” or not, since I had not created the original files. The AI revolution takes that idea and pushes it about 100 notches further, in my opinion.

I’ll keep an eye on this field of AI Music, as songwriter, musician, fan.

Peace (and Sound),
Kevin

PS — Update — I wanted to try out making a song via AI, so I used UberDuck’s tool to create this rap: https://app.uberduck.ai/rap/song/6908cc42-58d3-4f3f-8411-5f3faf844270 

I can’t say I am all that impressed but it was interesting. I also used the free account, which is clearly limited in many ways.