Re-melodization of a Tune

Re-crafting a melody and more, using UDIO.

Jun 29, 2024

Introduction

I have written a bit about recrafting a tune by revising the melody, harmony, accompaniment, and overall vibe (see below). I have gotten results using the music AI at UDIO.COM (soon to be bankrupted by the music industry majors through copyright lawsuits, I think).

I have taken one of my tunes and come up with six UDIO versions. The tune was called “I'd Be a Fool to Take You Back.” Each version uses the lyrics from my tune.

I think the results are pretty good in all cases, but what I have found is that the melodies, phrasing, and emphasis are quite different in each case. Previously, I had considered re-melodization as an exercise for a person, probably via singing extemporaneously over lyrics, or maybe lyrics plus chord progression. With UDIO, I get some quite dramatic results from an algorithm, running on a database scraped from existing tunes!

So, if I go to sing the tune, which version do I choose? Confusion reigns in my mind; I can not decide!

Problems with UDIO

UDIO is amazing, but I fear for its future. However, copyright aside, it has a number of annoying features that temper my amazement at its capabilities. I may not fully understand how to drive it, so take my observations as provisional.

It produces random results, slightly corresponding to prompts.
It ignores prompts.
It rewrites prompts (although there is a control for that).
It substitutes its own prompts for names of artists, groups, or tunes. Sometimes these are apropos, often they are not. Presumably, this is done to avoid copyright cases. Ain’t workin’.
It will only accept a small amount of your lyrics at present, even with a paid subscription.
It can go off into the wild blue yonder and ignore your lyrics, add its own, or add absolute gibberish.
It sometimes goes off into lyric hyperspace with added material and noises, often unintelligible, and it is hard to bring it back.
I does not respect your intentions as to where the music pulses are, and there is no way to convey such things to the algorithm.
It adds random pauses and fills between lines of lyrics when such are not indicated.
It will often add endless fills and instrumentals, intros where none are desired.
You cannot seem to get a voice reliably; it is hit or miss.
It adds extra voices to what should be a solo part. Sometimes it adds a pair of voices.
You cannot change the voice on a tune once it is established. For instance, my intention on “I’d be a Fool to Take You Back” was to have the bridge done by the distaff partner. It makes more sense that way. However, I have not been able to change the voice to another. No dice.
Overall, there is little control over the output; no direct control.

So, now, for my tune:

UDIO Does "I'd Be a Fool to Take You Back"

Adult Contemporary Maybe

1×