Is 'programming fatigue' a real problem for you when songwriting?

[JamAi]

AUTO-ADMIN: Non-MP3, WAV, OGG, SoundCloud, YouTube, Vimeo, Twitter and Facebook links in this post have been protected automatically. Once the member reaches 5 posts the links will function as normal.

Hi KVR community,

I've been a longtime reader and have a concept I'd love to get your honest feedback on.

Like many of you, I'm a musician first and a producer second. My biggest creative bottleneck is often that moment after I've found a cool riff or chord progression on my guitar/piano. The momentum dies as soon as I have to switch gears into 'producer mode'—programming MIDI drums, finding a bass sound, etc. I call it 'programming fatigue'.

To try and solve this for myself, I'm exploring an idea for a tool called JamAI.

The concept is simple: it's an AI that listens to your playing and generates a fitting bassline and drum groove in real-time, almost like a virtual session player. The goal isn't to replace the producer, but to act as a high-speed musical sketchpad to keep the initial creative idea flowing.

I put together a simple landing page to visualize the idea: https://jamai.carrd.co/ (https://jamai.carrd.co/)

My question for this community is: Does this resonate with you? Do you experience a similar workflow problem? And do you see a tool like this as a helpful creative partner, or just another gimmick?

I have a thick skin and am genuinely looking for your expert opinions to see if this is a path worth pursuing. If the idea seems interesting, there's a beta sign-up on the site.

Thanks for your time and expertise. Ed

[Tracewidth]

If an AI agent wanted to pay me for the opportunity to study my technique I might consider it, if the money was right.

[jancivil]

EDIT: I wanted to show the DAW workflow as conducive - I'm not stymied by searching for sounds, I have a process - but I went off on a whole trip.

I spun off from the premise in the topic title. The rest was Tl;DR. Turns out it doesn't really address what it purported to set out to do. It's about promoting an "AI" to provide eg., bass parts. Oh well. I say something below about it.

f**k AI.

[jancivil]

as per "tool like this as a helpful creative partner" I wouldn't consider AI - which is a super-powered averaging engine instead of a thinking apparatus - capable of a genuine response. It has no experiences in a world, no emotional history, if it seems to have a style it's again an average constructed purely out of other [human, if it's looking at music for real] experiences. It doesn't think eg., 'this is good', that's an average drawn from its database. It can be seriously well trained to find a good instance of what there is, for what to do, but it's not a creative partner as I know it to be.

So where the goal is find a generic bass part for a kind of standard chord progression one supposes this can work. (actually I'd have high confidence of that)
I don't do that.
But the difference between me and someone that feels they need a bot partner to feed them a bass part is I LIKE WRITING PARTS. I'm good at it.

[fatprice]

No because while I cannot speak for all songwriters it is the beat and bass that begin before the melodies 9/10 times.

[wagtunes]

In think a lot of songwriters would benefit from this tool. I wouldn't use it myself but I think there is definitely a market for it.

[lfm]

The concept is simple: it's an AI that listens to your playing and generates a fitting bassline and drum groove in real-time, almost like a virtual session player. The goal isn't to replace the producer, but to act as a high-speed musical sketchpad to keep the initial creative idea flowing.

My question for this community is: Does this resonate with you? Do you experience a similar workflow problem? And do you see a tool like this as a helpful creative partner, or just another gimmick?

I have a thick skin and am genuinely looking for your expert opinions to see if this is a path worth pursuing.

I think you immediately run into selecting "styles" and maybe "genre" as well.
- all arrangers including BandInABox work like that
- then you are in the territory
- hard to have AI really hit the nail on the head?

Let's say you get one song you are satisfied with
- what about the next one?
- how long before they all are somewhat similar?

I would soon feel like when you come to a telephone service that absolutely want you to voice what you want
- "I did not understand that"
- and frustration steems up

25 years ago I used a midi based JammerPro that were similar.
- so with a GM library you could get something
- pressed a key and it changed style until you were close enough
- if you had a melody it could suggest a chord progression

Pattern based drum machines is probably the most creative thing to get a foundation of some sort, for the beat at least.
- hitting kit keys in realtime, holding the same key down to erase if not good
- you can find "little accidents" that are cool
- then you copy that pattern and make variations with fills and stuff
- chain together patterns to a song

But making a riff or chord progression you probably have something in your head while doing that.
- that AI would sense that?
- press a button and it suggest a new beat+bassline, repeat until done
- lock beat, but vary bassline or vice versa
- maybe something to play along with but hardly final version

Way to go for me was to first do keyboard drums, doing realtime like I would on a drum machine.
- then pads
- the eDrums and learn to play drums too
- not for everyone of course

If you have a guitar and play that, use an octaver like POG2 and get a line on guitar octaves down to make the bass line.
- in real time

Just use loop recording until in the ballpark.

What you try to do is really hard. We all struggle with this.

I think it was Roland Jupiter X/Xm that the sequencer did adapt and do something along your playing. Maybe there are demos of that how good it is.

[Ah_Dziz]

I wouldn't use the proposed tool. If I was so burnt out that I couldn't come up with anything I'd just stop before I'd use a machine learning algorithm. I generally haven't had that problem which is why I can work to deadlines etc.

These types of things seem like they might be slightly interesting as auto accompaniment to practice playing live if they were capable of such a thing. I'd never ever use it to write a song though.

[jancivil]

The concept is simple: it's an AI that listens to your playing

But making a riff or chord progression you probably have something in your head while doing that.
- that AI would sense that?

AI does pattern-matching, using machine language that's a total black box to a human. It is not listening, it will search its training and a database for a match for your audio. To the extent what you'll do is totally generic and has been done a million times its chances of spitting out a copy derived from its sources will tend to be good. It isn't jamming with you.

No, it has no sense of thought or of a moment in time, it has literally no awareness.