Generative AI startup Uncharted Labs launched Udio, a text-to-music generator that generates high-quality music from text prompts.
A few weeks ago, Gen-AI enthusiasts were gushing over the music Suno V3 generated. User feedback from early testing of Udio suggests that Suno V3’s time at the top may have been short-lived.
Udio is like ChatGPT for music. You prompt it with a description of what you want the song to be about along with a musical style and hit the perhaps ironically labeled “Create” button.
Udio auto-generates lyrics to go with your new creation but also allows you to write your own lyrics in English, Chinese, Japanese, Russian, Polish, German, French, or Italian.
In a few seconds, the tool generates a 30-second clip that you can extend to 90 seconds and edit on the platform.
Tracks generated by early AI music tools were fairly generic, repetitive, and pretty basic. Udio’s musical and vocal output is complex and sounds as good as a lot of the commercial music produced by human artists.
It’s not going to have you abandon your Pink Floyd or Led Zepp albums but the fact that music like this is completely AI-generated is wild.
Crater Lullaby pic.twitter.com/2jZGv7TfIw
— David M. Comfort (@DavidmComfort) April 10, 2024
Training data and fine print
Uncharted Labs’ founders include former Google DeepMind engineers. The company raised $10M in seed funding from investors including Andreessen Horowitz’s a16z and Instagram co-founder Mike Kreiger.
will.i.am of Blackeyed Peas fame is one of the notable musicians who have invested in Udio. He’s been outspoken in his support for AI music, a view not shared by the 200 musicians who signed an open letter last week decrying AI’s involvement in music production.
Udio’s creators haven’t divulged the details of their training data but the tool was almost certainly trained on copyrighted music.
Udio’s website says that “Udio doesn’t generate songs using artists’ voices”. That may be, but it sounds like the model spent a lot of time listening to human artists.
Rolling Stone’s Brian Hiatt noted the familiar sounds of Tom Petty’s voice in some of the generated songs.
Udio, the new Suno competitor, twice generated songs for me using Tom Petty’s voice. https://t.co/DJxnVjQaNT
— Brian Hiatt (@hiattb) April 10, 2024
The fine print in Udio’s Ts&Cs urges anyone aggrieved with their service to follow an “informal process first” rather than going the legal route. I suspect we’ll see more class-action copyright lawsuits soon, similar to those OpenAI, Meta, and Microsoft are embroiled in.
Udio is free during its beta testing phase and allows users to create up to 1,200 songs per month which they can use for commercial purposes.
If you make a catchy tune using Udio, bear in mind that you still can’t copyright AI content in the USA. And you may feel a little conflicted when you tell someone, “Hey, listen to this song I made.”
