Suggest chord progressions

HI I do not know if Scaler has this already or it has been requested. Can we have scaler suggest chord progressions like 1-4-5-3 etc and supply the chords based on chords we have already selected in section C?

Hi @xtremsounds

we are doing a couple of things around this space. Version 2.4 will have new content with “starter” progressions, this will allow faster bootstrapping of new ideas from simple starting points.

Regarding chord suggestions based on already existing content, we are working on it but it is hard to make things interesting and compatible at the same time. The “next chord” is a really vague concept, lots of things could fit, it really depends on what you intend to do. It also depends on the amount of content you already have.

For example, if you try to embellish an existing progression, we might duplicate some chords, vary some voicings, use inversions and add obvious passing chords. This will keep the main idea behind the progression.

But you could also be trying to complete a progression, find a B part, maybe you want to take it in a different direction using out of key chords, modulation etc… in this case pretty much anything could be used.

Do you want to share a bit more about what you would like Scaler to do?

Our goal has always been to show the theory behind what Scaler does and we try to avoid turning it into a preset box you click until it sounds good. We will get there, and as usual, thanks for the feedback :slight_smile:

Thanks. What about providing a list of chord progression options that when we click that it just rearranges based on our selection? That might be easier.

Or maybe some formula like this will work? I found a reference here

I next possible chord Anything
vi next possible chord IV V I ii
IV next possible chord I V vi iii
V next possible chord I IV vi
iii next possible chord vi I IV ii
ii next possible chord V IV iii

1 Like

@Bernd question to you as professional Data Scientist: The „next chord“ could be understood as a prediction task. I would imagine an AI (weak AI) module where scaler would provide multiple suggestions paths on the basis of current chords selected in section C and an appropriate model (maybe it might be split by genre). The user may then either take over a suggestion or manually overide which would retrigger a new prediction iteration, and so on.
Any thoughts?

In theory this is doable, but “prediction” in the data science sense is based on a training data set. AI learns from patterns observed in large sets of examples given. So what chords are a “natural” progression to a given chord highly depends on the musical styles the model was trained on. For this to be useful in Scaler, for musical composition, you would either have to allow Scaler be trained by the user with example music sets, or Scaler to come out of the box with pretrained models based on established music styles. Both of the scenarios are already having a good foundation in Scaler by virtue of Scaler being able to take in MIDI files, and providing a library of music genre based chord progressions. If there is serious interest in this feature, it is definitely doable but a major expansion that would probably take another 2 or 3 major releases of Scaler to become a mature feature. Is there a product out there that does anything similar to this?

@Bernd Let me disclose that in my day job I am a Data-Scientist too, and hobbywise during Corona got interested into music production (having amost no music theory background nor any talent).

My primary interest is to experiment with music tools like scaler and enjoy the tonal emerging results by toying (to be honest lets call it nerding) around with parameters, rules, modulation and automation.

I would be very interested to develop the idea further of a ML-based progression module inside or outside Scaler and would even go so far to contribute by offering algorithmic help if an appropriate working mode can be found.

No, because if you make one good enough, you simply release the generated music under your name and don’t tell anyone :stuck_out_tongue:

On a more serious note, we are looking into it, but to be honest most suggestion engines give pretty average results. And there is the question of the value it brings to users… If you have to generate 7, 8, or 20 results before you hear something you like you could just click around in Scaler at random.

We are planning a few things in this space for 2.5 but I don’t think it will become a main focus of Scaler, we try to help people make music faster not replace them entirely :wink:

Thanks @Ed1 for response.

I second you by saying that a Scaler user still should be in the driver seat. I was more thinking about offering fast musical ideas which are picked by users taste.

By the way there are ways to even algorithmically seed some kind of element of surprise to beat the boring aspect which I also see when using ML only.

I think things should be kept simple such as having some patterns to choose from and apply them to the choice of chords. That would at least get some things started. Of course after finding that chart from The Shed website, I can just refer to that if I need it.

I’m with you, @Ed1 , on both accounts…

No, because if you make one good enough, you simply release the generated music under your name and don’t tell anyone

I especially like this stance…

we try to help people make music faster not replace them entirely

@Cliff, like I said, it would be fairly straight forward to feed a machine learning model existing songs for a particular genre to estimate the likelihood of what chords would come next, given an existing chord progression. This could be helpful if someone would like to write a tune for a particular music style. I have concerns with this scenario, all of them are human not technical concerns:

  1. This likely would lead to more music produced that sounds similar to existing music, minimizing happy accidents that lead to musical evolution
  2. It would condition new musicians to listen more to a computer, and even if not right out automating making music, it would lead to a similar syndrome of “autopilot” flying, why commercial airlines are forcing their pilots to more manual flight hours on long flights. Ultimately skills would erode. I use Scaler because it helps train my musical skill. I don’t expect it to do work for me.
  3. Instead of training a machine (“machine learning”), I would much rather see enabling more people to learn.
  4. Instead of “predicting”, you might end up having computers/software “prescribing”, because people would get into the habit following the computer’s output, instead of making autonomous bold and creative decisions that lead to innovation and variety. This is a distinct argument from point 2.

I am of the viewpoint that technology should be used to solve a pressing human problem, and not just for its own sake. And much of the hype around ML and AI that I witness is focusing on what to do with that new tool. I have a hammer, and now I am in desperate search for nails to pound in.
So as a product manager I typically look from the other side, what is the actual human need, not for a particular feature, but for the human to flourish, feel empowered, enriched.

A completely different way to take on the challenge, “how do I make better music” would be to have a social voting system in place, a bit like supervised learning in ML. You have the computer produce chord progressions and related performance expressions, and let real people listen and vote on them. This way you would get better feedback not on what “typically” comes next, but what actually sounds good/interesting to people, especially if you are a professional musician.

I am a firm believer that humans should always be kept in the loop, both for moral as well as quality-of-result reasons.

1 Like

@Bernd I think your different take is interesting. But my intuition is that training with already existing harmonies from professional musicians will lead to similar results, as good sound is very much cultural common sense. Anyways thanks for dicussion.

Now that is a very curious assertion. Very different take indeed.

What culture? National culture? Subcultures? How come we (humanity) have so many scales for music? How come there is micro-tonality? How come musical genres sprout like mushrooms ever since most national societies have turned into liberal democracies, with better representation of individual preferences? (before that, “common sense culture” told you what is appropriate to like). How come “we” (human culture as a homogenous assumption) once declared heresy the tritone, jazz, the Hendrix chord, punk, or making music with a computer? If you look closer, it’s usually an established social group (during my growing up it was the adults who “decided” what is music and what is not). It’s the classical argument over what makes music music, let alone “good” music.

It seems to me that the major social consensuses amongst sub-culture tribes, e.g. people who like jazz, or blues, or classic music, or neo soul, funk, you name it, is already reflected in Scaler via its organized chord progression library by genre and artist, and I can make use of that today. Of course the library of chord progressions could always stand expansion, and I am in favor of an open source ecosystem where musicians more easily can directly contribute to a cloud-based Scaler chords library that becomes instantly available for each Scaler user. And Scaler’s myriads of scales representing many cultural traditions, across various ethnicities, tribes, geographies, instruments, are also already there and browsable. So the immediate opportunity to make that “common sense” recognizability (e.g. within a given genre, artist style, or scale) happen with low friction as a musician is in the user interface improvement, no need for ML or AI for this use case.

Why am I so passionate about this argument? My parents, my school teachers, my “culture” (old school Germans, leftover Nazis), constantly told me as a young person that I am “unmusical”, “untalented”, and my innocent childish experiments that I so proudly showed around were “awful” - just for me to find out later that I was well in sync with punk, rage rap, grunge, and nordic viking metal :slight_smile:

I have a tendency to use chords that “cultural common sense” (that is in the tradition of western imperialism) are considered “sad”, “melancholic”, “dark”, yet me and my (sub-) culture of “whatever-you-wanna-label-us” find them cathartic, relieving, bringing us closure and acceptance. You see what I am saying? And Ed’s been saying it for a while in this forum, as an explanation to feature requests for “better” chord matching, to paraphrase him, how musical “common sense” (perception) is in the eye of the (cultural context) beholder.

I think the bigger opportunity than to create a tool to “fit in” by sounding compliant to mainstream notions, is a tool that empowers, creativity, confidence in one’s potential, tolerance for being different.

1 Like

I’m sort of with Bernd’s take on this one.
I tried Captain Chords for some time, which does have a large selection of ready built progressions to catalyse some tune, and have also tried some of the midi bits from DMS. I felt a little frustrated with this because it was very much a ‘cookie cutter’ approach; although I might have generated a few Trance pieces, they sounded like every other Trance piece (maybe that’s just the nature of Trance…)
Then along came Scaler and I have not used either since. Even if I start off with one of the available progressions, the next port of call is the modulation page to explore and experiment. I may produce crap, but it’s my crap, and hence not only do I feel more creative, it furthers my interest in learning more about music theory.
I’m sure Davide must have had giving folk the facility to explore with their ears as one of his core goals.

For me personally scaler should assist our creative process not dictate it. If it gets in the way it’s a hindrance to creativity. If it holds your hand it’s a hindrance. So ultimately it would be even more seamless with no roadblocks or dead ends. The whole point of scaler is to make it easier to make the music that is naturally in your head. Hence the ‘empower the composer within’ tagline. There’s a fine line between that and suggesting / recommending / sending you on a merry path. Too much other software out there will do that better than Scaler.
We have some very exciting directions for scaler. We are still finding that balance but the path ahead is clearer and more obvious.


Interesting philosophical path!

I think I am close to @Cliff with the meaning I love HAL9000 suggesting me chords, melodies etc, but on the other hand I don’t want HAL9000 playing music for me because it could be boring, and I love Scaler because it is balanced: not to much hassle, not too much suggestion

I don’t know if I explained myself well…
My ability to write English is not enough to fly so high as @Bernd

Nah, I just have Scaler-for-English, a tool that helps me create creative, long-winded, meandering, convoluted phrases, in English :wink:

1 Like

LOL I love the brit… ahem the German humour

That’s the problem…this joyously addictive journey goes on and on and on and next thing I know it’s dark outside, the snow has melted and I’ve not cut the grass. :slight_smile:

1 Like