Category Archives: Seeded Tunes

Seeded Tunes – Scene and Labels

A ‘scene’ is a nebulous concept, but it’s pretty well understood.

David Byrne in How Music Works talks about the necessary ingredients for a music scene, including location and venue(s), with a description of the old and new CBGB’s in NYC.

Julian Cope in Head On describes the nepotistic maelstrom of Eric’s in Liverpool in 1978-9. Location, people, financial circumstances, common beliefs.

In the case of Seeded Tunes, the scene is the root of the whole output, and it basically consists of a pool of fictional people, and a random number of record labels. The ideas there is that these people got together somehow, and from that outpouring of creativity came artists of course, but also labels, which I don’t understand, but they seem important, especially in our digital future as a form of curated marker of “you might like this”.

When the scene is created, a number of fictional people are generated – the “scene pool”. From there, labels are created, each with a number of people drawn from the pool. It’s possible that the same person might be involved in more than one label.

Each label therefore has a “label pool” of people, and from each label a number of “artists” is created. I settled on the word “artist” rather than “band” because the latter seems limiting, somehow.

I’ll describe the artist process in another post.

Seeded Tunes – Random?

The lynchpin/keystone/root idea of Seeded Tunes is to use a random number generator (RNG) to drive the creation of text, images and music.

Regular RNGs are not truly random. They use stuff like the current date, combined with internal values such as drive space and other numbers to come up with something that resembles a series of random numbers. But the creators of these RNGs go to pains to make it clear that these are not truly random – someone with enough access, time, and skill could predict the numbers based on the provided seed. As a result, regular RNGs are not recommended for security-related applications, such as cryptography or finance. There are other more robust and rigorous options for those uses.

Leaving that aside, how much of the output of this project is truly random? How random do I want the music to be? Could I not say that truly random music would be a random string of bits encoded into an WAV file? That would probably be unlistenable – not in the aesthetic sense (I don’t care about that) but in the sense that it would not be a valid sound file.

If I channeled the randomness into making a valid sound file that could be listened to, would it be long? Short? A frequency outside human hearing? It became clear that true randomness would be no good for creating what I wanted.

But what did I want to create? If I broke down the actual components of the output, they fell into the following groups:

  • Names of people
  • Names of artists
  • Names of albums
  • Names of tracks
  • Artwork of albums
  • The music itself

I decided I wanted to create recognizable music, with realistic (if unusual) names, made by fictional people with regular (if unusual) names, and with artwork that falls into a particular graphical style I am a fan of.

That meant I had to reduce the amount of randomness in varying degrees for each component. Initially I felt that this was too much of a compromise, but in the end I think I’ll be happy with a realistic output, realistic names, attractive artwork, and music that was structured enough, but without following the conventions of music theory, or any of the methods that generative music uses.

It will probably be unlistenable – but who is to judge that? Bandcamp, Brainwashed, and power electronics showed me the way there.

Seeded Tunes – Why

I’m a frustrated musician, in that I would like to be a musician but I have been cruelly prevented in chasing this dream by A) a lack of innate talent and B) a lack of will or discipline to learn.

Sure, I’ve dabbled. There have been attempts to get some kind of music project going with friends, but despite their input and enthusiasm, the issues described above have always got in the way. As time has rolled on, and responsibilities have become greater, I’ve come to accept my limitations.

But as time rolled on, I also read about the clever ways people were using computers to generate art. I’ve always been a computer person, but I’d let my coding skills lapse. As I aged, I became more interested in minimal music, drone, electronic noise and the like. Being invited to write for Both Bars On exposed me to more varied stuff as well, as did listening to the Brainwashed Radio Podcast. Seeing and hearing the vast array of stuff on Bandcamp crystallized the idea in my brain.

If I can’t make music myself, I could write a program to do it. It wouldn’t need to be smart, or use any of the tools or techniques used by generative artists to make human sounding music. It would be truly random (for certain values of “truly” – see a future post about that).

If I could generate a track, why not a whole album? If I could generate an album, why not many?

Who is making these albums? What are their names? What label are they on? What year was this made?

Were there multiple artists making music? Did they know each other? Were they part of a scene? Did groups share members with each other? What would the family tree of this scene look like? Would Pete Frame draw the diagram, and would there be a TV docu based on it?

I was reminded of Bill Drummond writing in 45 about the fictional Icelandic underground he created. I thought of the blog For Those Who Tried to Rock, which chronicled the bands, gigs, dreams, and attempts at rock glory by small-town musos.

That’s how the project found its current form. A way to make up for time lost to inability, and share with the world multiverses of (admittedly similar) fictional artists, and their art.

I’m just the curator.

Seeded Tunes – An Introduction

For quite a while now I’ve been chipping away at a project that combines music, art, my fascination with music family trees, generative processes, and laziness.

The idea – Seeded Tunes – is to use a random number generator to create music. But it’s a lot more complex, and in some ways a lot simpler, than that.

There are many tools and projects where people are using computers to create art, music, literature, you name it, with random elements. There might be a Markov Chain driving the creation of a written article, or a semi-AI using a corpus of existing creations to generate something similar but completely original. My project is not like that, but has parallels.

In a nutshell, you give my program a number (the “seed”), and it will generate an entire “scene” of music, with multiple labels, each with multiple named artists, each with multiple named albums, each with album artwork and multiple named tracks, each with generated tunes and drum patterns. It will also create a “Rock Family Tree” broadly in the style of Pete Frame’s fascinating work, which I’ve talked about before.

There are many aspects to this, and I’ll describe them in detail over the next few posts. It’s all been rattling around in my head for too long, and people have taken an interest, so here’s where to get in the know.

In the next post, I’ll talk about the WHY…

Rock Family Tree Auto-generation

I’ve been fascinated by Pete Frame’s Rock Family Trees for a long time. I created a tree in that style for my friends and associates over on the Coiled Spring site, seen below.
Flange Circus Family Tree
I describe the process for manually creating the diagram over on that site, so check it out if you’re curious. I had other ideas after that, so read on.

I’ve been playing around with the tools I used to create this diagram. Graphviz especially. It’s a very powerful tool for automatically laying out diagrams. You provide it with a text input file defining the nodes (boxes) and how they are connected with edges (lines or arrows). The file doesn’t include any positioning information; Graphviz arranges the diagram itself, aligning nodes, routing edges, and making a nice neat diagram. There are various options for influencing how the diagram is arranged, but overall it is automatic, and very smart.

You can write the input file by hand, but for complex diagrams with many nodes and edges, it makes more sense to generate it using another tool. There are many examples with tens or even hundreds of nodes, generated by software analysing computer networks, chemical reactions, mathematics, and so on. You can see some crazy examples in the gallery here.

This lead me to think about the possibility of writing a program to generate a Rock Family Tree from data held in files or a database – and that coincided with an idea that had been rattling around in my head for a while. Why stop at the family tree? Why not generate the actual music and artists as well?

And thus Seeded Tunes was born.