AI Text to Music Production
My experience
The world is changing at a rapid pace. AI has finally really come into its own recently. One of the more interesting areas of AI as related to day to day life is the use of text to music.
For me, text to music AI has revolutionized video productions I’ve been working on on Youtube. Before, you could use the crappy stock music Youtube had to offer, or create your own music. I’m not a musician, nor do I play one on TV. So what I had been doing was using MIDI classical music tracks and modifying them to suit my needs.
MIDI if you didn’t already know, is basically like sheet music for computers. All the classical music you can think of is already transcribed to Midi files and is presumably copyright free if it’s over 100 years or so. So I’d been taking classical tracks from some of my favorite composers, for example both the brother and sister Felix/Fanny Mendelssohn and changing the tunes a bit for fun and using their music for my videos.
Then came AI text to Music.
I should start by pointing out that I’m not an expert on the subject. But by simple experimentation, I’ve come up with lots of useful music for my videos. And even some that I like to listen to like I would other tunes from real musicians.
Here’s my take on what’s available now.
There are two big players, Udio and Suno. I subscribe to both for $10/month. That gives you a certain amount of credits. Honestly, I’m not sure how many credits it takes for each song.
What’s the same between Suno and Udio
- You can generate text to music songs
- They can be instrumentals or with lyrics
- Both services can generate lyrics or you can BYOL (bring your own lyrics)
Suno is my preferred of the two services. Use of their interface makes more sense to me. Suno generates songs up to 4 minutes long, and those can be extended. You can put enough words in the initial four minute segment (3500 characters) to make a conventional song.
Suno also generates songs much quicker. And while generating, you can start listening before the tune ends.
Udio I find less useful for me. With their service, each prompt creates a segment of music about 1 minute long. And the default segments don’t end gracefully, they can end in mid sentence or in an awkward musical place in a track. To fix that, even though I’d often be fine with a one minute long track for my videos, you have to extend the track (for more credits) and guide the song where you want it to go. Udio also puts some onerous limits on text length, so you have to break up your custom lyrics into segments to match your one minute allotment.
That being said, you can explore what other have created on Udio, and there’s some amazing work out there. My sense is that Udio may be like a manual car, or a camera on manual. For an amateur or beginner, it may not be the best choice. But if you know what you’re doing, you may be able to create something great. But that’s not me.
Here are a few samples 1, 2. I plan to start adding Lensbusters.com songs for download here as a regular feature soon.
Please note this was written in July 2024. Like me, this text may not have aged gracefully.