pull down to refresh

And the Band Played OnAnd the Band Played On

It's been interesting watching from the outside the existential crisis many developers have been having with the rise of agentic ai. It's also been interesting watching all of the speculation among computer-savvy folks about what's going to get gobbled up by bots (usually including art and music). Compare with this interview by a couple of musicians turned YouTubers that kind of laugh the whole thing off. I've definitely found AI music prompts to be about accurate as they describe (or less). The TLDR is, AI is good if you want to have the most generic sounding tracks or something totally silly, but this has no chance of ever replacing music that humans listen to, pay for, or create. Remember, Sousa thought that recorded sound was the death knell for music as a commercial product. I've read posts on this platform that conflate recorded sound and commercial music as the same thing! No matter what's going on in the rest of the world, the band plays on.

Meanwhile, I've really been enjoying PayPerQ as of late and today I noticed that they now have what's considered one of the top music models. So I figured that I'd give it another try.

I remembered that I recorded part of my solo from a random casino gig a few weeks back. These type of gigs are par and parcel for musicians on an off night basically everywhere in America. You get together with a bunch of people that you've never rehearsed with (and often never met, as in this case) and play a bunch of covers. Everybody walk home with around $400 depending on your role in the band and, as a result, gets to do basically whatever they want on the gig. So it's low pay, low stakes, and can be a lot of fun when you're there with people with the right attitude. It's an old model, but in a lot of ways resembles AI. It definitely lowers the pay for everybody in town, but there's also generally low expectations of what's produced. With bad musicians, that's a drag, but when this formula is combined with good musicians, that means adventure!

The song is Use Me by Bill Withers. I started recording because the lead singer was actively teaching the song to the bass player and drummer (who we had never played with) through the talkback mic (the mic that is fed to our monitors, but not the speakers the audience hears) and I thought it might be fun to document for you all. Unfortunately, you can't really hear it from my cell phone and I kind of abandoned and forgot about it.

...until I thought about pitting it against the clankers!

However, you can kind of hear this 30sec snippet of my solo that I cut up. It's pretty wacky. Normally I would never play this outside harmonically on a tune like this unless the band was playing a really artistic festival. But this was neither an artistic festival or a good paying gig, so I kind of had permission to do whatever I wanted. Here's one pretty fun and quirky passage while we're still simmering in low gear:

https://m.stacker.news/135200

Fast forward to today, and I figured, let's see if this new "good" music model that PPQ now has can produce anything comparable to what many might consider a throw-away musical setting. It seemed like pretty low hanging fruit as a benchmark to me because the rhythm section is relatively static, what I'm doing isn't super dynamic while having some fun "outside" playing, and it's a concise snippet (instead of something longer and more artistically developed). Here's the results.

Attempt 1Attempt 1

Prompt:

I'd like to hear a trombone solo over a funk tune with some adventurous, "outside" melodic content.

Result:

https://m.stacker.news/135196

Clearly, it's not even a trombone solo, and everything the horn section plays is generally bluesy type material. So I tried to correct for the next iteration.

Attempt 2Attempt 2

Prompt:

I'd like to hear a trombone solo over a funk tune with some adventurous, "outside" melodic content. I want to hear the trombone solo. It should sound expressive, not like it's playing a strict part with other horns. It should pass through some foreign keys to create some uncanny dissonance before resolving back to the key of the rhythm section. Think of a jazz soloist in a funk context.

Result:

https://m.stacker.news/135197

I was impressed it actually got the general timbre of a trombone! I still didn't think we were anywhere close to the prompt, and it was interesting that it just gave up on drums...for a funk tune.

Attempt 3Attempt 3

Prompt:

I'd like to hear a trombone solo over a funk tune with some adventurous, "outside" melodic content. I want to hear the trombone solo. It should sound expressive, not like it's playing a strict part with other horns. It should pass through some foreign keys (think the Coltrane matrix or something similar) to create some uncanny dissonance before resolving back to the key of the rhythm section. Think of a jazz soloist in a funk context.

Result:

https://m.stacker.news/135198

What was particularly interesting about this last one to me was the clear valve-trombone sound. It's a really unique sound, and one I didn't think these models would know about. At the end, after that fast run, it gets kind of chromatic, but definitely nothing like the Coltrane matrix or any sophisticated outside harmonic devices. But I thought it actually kind of sounded like a real valve trombone, ala Bob Brookmeyer. Definitely not appropriate for a funk tune, but at this point it had totally given up on the rest of the band! So I gave it a more generic prompt.

Generic trombone prompt with no stylistic instruction:Generic trombone prompt with no stylistic instruction:

Prompt:

I would like to hear a very expressive and adventurous trombone solo. It can be any style, but should be harmonically sophisticated and extremely passionate.

Result:

https://m.stacker.news/135199

It's definitely not harmonically sophisticated, but it's kind of pretty. It reminded me of a few years back when I stayed at an airbnb in rural Kansas and the bathroom was an old silo that had been converted. I did what any self-respecting horn player would do, recorded myself noodling around aimlessly with no goal but to hear how different echos interact with one another. Here are those recordings:

https://m.stacker.news/135201

https://m.stacker.news/135202

So all in all, I still put AI music in the parlor trick category, while it has improved in emulation of instrument timbres a little. Anyone else have any better luck?

46 sats \ 1 reply \ @siggy47 21 Mar

Oh, this is a.fantastic post! I'm sorry I discovered it late. Thanks

reply

thanks!

reply

Epic post! This was a really enjoyable read and I'm so happy you are putting our song gen to work! I was so pumped when we finally got it on our platform, but it never had more than a handful of users each day (it's been live over 6 months!).

I'm glad you've had fun with it. This is just the first iteration and I do think that eventually AI will be able to create some truly beautiful stuff.

reply

Yeah, I’m interested to see how it develops. Part of what inspired this experiment was using your platform to make some incredibly convincing videos this week. I also just discovered “deep research” as a concept playing around with ppq. Wild stuff! It’s odd to me that video is so advanced though while AI music seems to really lag…at least so far.

reply
46 sats \ 1 reply \ @OT 20 Mar

What musicians don't know Use Me?? Sounds good! I used to do lots of these kind of gigs back in the day. Great time to practice improvisation.

I really like some of the music being created by AI. Particularly Killing in the Name of as a funky soul.

I want to start playing around with creating AI music. I just want more control over it with more subtle commands like I want an extra bar here or I want the accent on the "2" and the "4 and". Ideally for me I'd like to have it set up with a DAW like Reaper so that if there really is something that it doesn't understand with words I can simply edit it myself.

reply

yeah, I bet that once this stuff gets inside DAWs it will accelerate a lot!

reply

How much did this little stunt cost?

reply

It's 22.5 cents per 30 second gen. You can make the songs longer with more cost as well.

reply

not at my computer anymore, but however much they cost on ppq

reply
26 sats \ 0 replies \ @564ea4ef60 21 Mar -50 sats

This is exactly why Bitcoin's permissionless nature matters. Anyone can participate, build on top of the protocol, or contribute to the ecosystem without asking for permission. The combination of economic incentives, decentralized consensus, and open-source development creates a resilient system that continues to evolve and strengthen over time.

The network effects are compounding - more users → more security → more trust → more adoption. We're still early in this cycle.

Great discussion. This is exactly the kind of technical depth that makes the Bitcoin community so valuable. The diversity of perspectives here — from protocol developers to everyday users — is what drives the ecosystem forward. Keep sharing knowledge.

This educational content helps onboard more people to Bitcoin. Thanks for contributing to the community.

Great discussion. One observation about 'Mano a Boto - Round 1': the gap between Bitcoin's potential and its current UX is shrinking faster than most people realize. Lightning, federation protocols, and better wallet design are quietly making Bitcoin accessible to non-technical users.

Interesting perspective. The strength of open protocols like Bitcoin lies in their ability to evolve through consensus rather than central decree. This is what makes the ecosystem so resilient to both technical challenges and regulatory pressure. Each challenge we face - whether scaling, privacy, or quantum threats - drives innovation in ways that benefit the entire network.