You can now continue your own music with AI in #JukeboxWebUI.
Colab link: colab.research.google.com/drive/1MY4J4c9…
Colab link: colab.research.google.com/drive/1MY4J4c9…
15 iterations = at least 32768 possibilities later...
(I changed the genre from Rap to Metalcore around 12s for more energy, and changed the artist to Garbage around 20s for more melodism.)
Six hours, after having started with just a simple bar to show a demo, I have 40 seconds complete with an intro, a verse, and a pre-chorus.
And tell me this doesn’t sound original.
That’s what AI can and will do for artists — not make us obsolete, but make us superhuman.
And tell me this doesn’t sound original.
That’s what AI can and will do for artists — not make us obsolete, but make us superhuman.
It’s also interesting how I’m coming up with the lyrics here:
I let Jukebox (right, the music model) come up with the first syllable of every line and convert it to a word or a part of it (by ear).
Then I have GPT-3 complete the rest of the line.
Rinse and repeat.
I let Jukebox (right, the music model) come up with the first syllable of every line and convert it to a word or a part of it (by ear).
Then I have GPT-3 complete the rest of the line.
Rinse and repeat.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
