Finding chords to fit a melody

Thanks Jamie! This isn’t meant to be “real music”, just an example of mostly unrelated chords that still work together. I would just want to find some way to enter the four notes and in return get lots of possible chord progressions. But I suppose that something like that just does not exist. At least not yet!
Excuse my upload-file ranting. I never upload stuff to forums so I was unaware of this limitation.

1 Like

Don’t forget that it is not necessary nor often desirable to have a new chord for each note of the melody.

It’s a great discussion and as I have mentioned before this is scaler’s raison d’être. I spent my young life as a successful touring music producer with no exposure to music theory. I was able to write chart melodies, riffs etc but mainly due to the genres of music I made, western harmony was not relevant, as I got older and doing more ‘composition’ Harmony became fundamentally important. I always wondered why there was no software that could listen to what I was doing and suggest harmonic content.
What the OP is suggesting is there but it is not in a ‘menu driven’ way. There are many ways to do it much like what @jjfagot suggests. Yes for me harmonic assistance and suggestions is the way forward in scaler and we have lots of that coming but how exactly can you better suggest chords to go with a few notes? I do like the idea of right clicking a note or series of notes in a melody and suggesting chords which share that note, ideas like that are practical.
For now I always suggest people look at the following sections which throw up nicely coloured chords that can not only work with your melody but suggest leading lines of where to go next.

  1. Voicings Menu (artists interpretations of diatonic chords)
  2. Suggest Mode (chords that go with not necessary belonging to scale)
  3. Circle of Fifths / Chord Edit (all available chords)
  4. Common Chord Sets in the Song Menus (thickened and coloured chords that go with)
    5 Modulation (chords that go where you may not naturally go)

If anyone has any practical suggestions of how a more structured suggesting chords to a melody would look I would love to hear them.


Davide, I absolutely agree with what you and jjfagot says! And the five methods you mention are indeed very useful and inspiring. And available now! No about about that. Did you have a chance to watch Paul Thomsons video? I guess I’m looking for a virtual version of him. Software with deep knowledge of music theory combined with a clever (AI-based?) “musical ear”.

  • I supply a melody (and maybe decides on some parameters, like how often should the chords change, etc.)
  • The app analyses the melody and suggests a progression of chords based on music theory
  • The AI (“Paul” :grinning:) decides what sounds “good” and what makes musical “sense”.
  • Repeat this a number of times until a “nice” progression is generated
  • I can then refine the progression further to my liking.

The Paul-step could initially be omitted and instead let the user listen to every progression and decide what to do.
I know that all this can be done manually with great assistance from Scaler but I’d love to see the process automated. It would save time and probably come up with progressions that are not obvious.
Such a feature can of course become quite complex but maybe not impossible to implement.

With all due respect I don’t think any AI is going to replace Paul Thompson’s experience and musicianship. Or at least not anytime soon or fully.

I saw Paul’s video and it’s a brilliant way to go about coming up with some ideas. But what you’re discovering is that is does require some learned skills. Dare I say that you probably just need to buckle down and practice and those skills will come. Scaler can certainly help with that by making some of the basic chords more familiar to you. But none of the AI tools will completely replace having some knowledge.

There is also nothing stopping you from using Scaler to find those chords and assist with interesting ways to voice them. It’s just not going to do it as quickly for you as Paul Thompson can do in his head. You may have to step through some menus and pages and yes… it will take you longer. Or I assume it will. I know it would take me a lot longer than it took Paul to sort out all those ideas. I would suspect his video was also edited as well. :slight_smile:

I think patience is the key here. Take some time rather than looking for the quick solution and the end product will probably be better as well. And then the more you try it the faster you’ll get at it because then you’ll know all those chords and how to quickly voice them without having to think too hard as well.


I agree totally although I do think Scaler can serve up options allowing you to choose and implement a little easier

I did watch the video and it is cool. This is all about imparting feel and knowledge, we all have feel but not all of us have knowledge - this is where Scaler comes in and I agree we can help. I’ve had a good think here along with this other thread: Chord Suggestion Based on the Top (Melody) Note and I think we may have something.

Thanks for all feedback as always.


Thanks RBIngraham,
I absolutely agree with you that I ought to learn more music theory and that patience is key and that the tools in Scaler are already great. No doubt about that! And I don’t expect AI to replace Paul anytime soon. I would happily settle for a mini-Paul Light. And I don’t want this because I’m lazy, I just think a system built on “actual” AI (not the marketing hype type) would be fun, inspirational, useful and very nicely complement the tools already present in Scaler.
AI is already way more capable than many people might think. A great and recent example is Google Imagen! You type a sentence and it creates an image! To me this is beyond mind blowing. And I believe that doing this with images is way more complicated and advanced than suggesting chords since with images there are no rules, like music theory, to start with!

Yes. It’s been quite a few years since a machine defeated a world chess champion. And yet, chess players use these machines to practice and learn, to simulate the results of their moves. Can you imagine that the chess championships were between machines, instead of between humans? Well, maybe there are already chess championships between machines, but it would be terrible if we humans stopped participating.
AI is making a lot of progress in all fields, and in music too, of course. I personally have many plugins on my PC that generate chords, arpeggiates, phrases, melodies, etc., but,
Let’s say that like other things I do with the help of machines, I finish my compositions “by hand”.
I think that Scaler and the other plugins help a lot by suggesting what I can do, but in the end the chord that I write in the score is up to me
By the way, I also liked this AI:

You write a text and it returns music that you can download in midi


You can also approach Scaler to study the chords so that you look at the tempo notes of the melody it contains. You then compare to Scaler chords or scales.


Well, I thought the most appropriate text would be

"Scaler is a composition assistant which identifies the scales of your tune to help you build powerful chord progressions and melodic sequences. Built to keep you in charge of the creative process, it helps you compose in a faster and more efficient way. Scaler also comes complete with hundreds of chord progressions and musical performances made by world famous artists across various musical genres.

Developed by music industry professionals and refined by a team of producers and artists, Scaler2 defines a new efficient workflow and helps you leverage the power of music theory in your own production. "

i.e. the first 2 paragraphs of the Scaler manual. It generates a type 1 3 track midi file; track one is comprised of chords, 2 of a single note around C0 (!), and track 3 individual notes with some very marked intervallic jumps. It’s not going to get a Grammy or get much airplay, but it’s the sort of thing @ClaudioPorcellana might like :slight_smile:

Anyway, I fed track 1 back into Scaler (Scaler analysing music produced by it’s own manual, eh ?) .

The chords were mainly 9 notes (d’oh!) and were too widely spaced for one patch to make much sense so they need splitting (some notes were repeated).


Running a few performances over them managed to generate something which sounded vaguely musical .The first 6 chords were interpreted by Scaler as
but I guess it stripped the octaves and duplicates.

I’m tackling something else right now, but I might just come back to it.


Introducing the complete Scaler manual would be much longer than a Bruckner symphony, hehehe
Check out this video from the show’s author:
LangoRhythm, every speaker is a musician | Jan Willem Kolkman | TEDxMaastricht - YouTube
In the end, he adapts the phrase very well, interacting with the interpretation given by the program and its rhythms and live effects.
I find it a bit hard to get things efficient, but it’s an AI in development.

1 Like

Here is an example of re-harmonization of a melody. It is a video by David Bennett. I like his videos, not too long, straight to the point and I always learn something. Sometimes something unexpected.
Maybe a future version of Scaler could generate something like this…


I’m making an AI, :smiley: :smile: :grin:
I have it in Reaper and now I’m getting it to work in Ableton.
I sing, I play the bassoon, or I just speak into the microphone and create music.
I leave here an image for you to experiment, if you want
At the microphone input I place a free plugin from Melda Productions (MTuner). By default, MTuner detects the pitch of incoming audio, but if you turn MIDI ENABLE on, it sends that detection wherever you want. I send it to Scaler, in which I have placed some chords (in A minor in this example). In turn, Scaler sends to a Pigments preset (you can choose any other instrument, whatever you like). The result is, at least for now, quite surprising. Try it, Yorkeman


An aside on AI (boooooring…)

Years ago, one of the fads of the software world was ‘OO’ or ‘Object Orientation’… Big system buyers were then (maybe still are) like sheep, and found comfort in buying what their peers had bought.

Hence, if you had ‘old’ software, then you ran the risk of being badged as ‘legacy’, which was sort of one step up from declaring at the dinner table that you have syphilis.

So everyone would suddenly start declaring their product to be ‘OO’ even though I knew some of them were just vanilla COBOL, or whatever. This slightly hacked me off, having spend 15M USD in building a ground up OO system.

Well, ‘AI’ is today’s ‘OO’. Vendors have to claim their gizmo uses AI to be modern and with it, and just like in yesteryear, I suspect that many of them have little of what a knowledge engineer would class as such.

AI is a broad class of techniques, and sits alongside related /overlapping technologies such as ‘Machine Learning’. Both are used by various previous and prior music apps (along with all those that don’t have an AI bone in their body, but declare so in the sales bumf.)

I think we are a long way off from emulating humans. Chess software is just mathematics. Software can generate music “in the style of Mozart” because it’s been taught or learned the patterns of the music and stuck them together is some random way, but constrained by rules - just as Mozart himself did in 1787.

But it will be a long time before you can give the machine a book on music theory and have it create something - there is, IMHO, something about human sentience which (Terminator notwithstanding) is hard to model with chips - it’s called ‘genius’.


I will … but of course you are not ‘artificial’ :slight_smile:

1 Like

You built a system that cost $15 million USD? What did it do? We’re you perhaps at one time a Bond villain?

This is excellent! Thanks for sharing.

1 Like

Investment Banking. Entry level licence was USD 1.4m plus project work and client mods. That figure was just the spend to the initial implementation in 2001 (it started in 1995). Its had vastly more spent on it since then as the scope expanded, and still going strong. You can see why I can never get my head around the price of Scaler…

1 Like

Hmmm. Well a Bond villain would have been more interesting. Carry on.

1 Like

Inspired by the David Bennett video posted above, I tested some alternative chords to the Eleanor Rigby melody. Did not use Scaler or any other help, just my ear. Maybe this sounds painful and dissonant to others… But at least I had fun! :grinning:
And this is what I hope a future version of Scaler AI can do in a few seconds! :grinning: