Hello from Japan! I discovered TrainJazz this morning and enjoyed it with my morning coffee — the idea of turning subway movements into music is quietly beautiful.
I would love to see a Japanese version someday. Tokyo’s train network is one of the most complex in the world, and I imagine the music it would make would be extraordinary. Thank you for creating something so thoughtful.
It would be cool to have multiple selectable cities, and compare them like that. I suppose to do it 'fairly' though you'd want to offset them so the peak times etc. lined up, so they all had a commuter rush at once. (London would be buzzing as I write, but the OP is pretty quiet – middle of the night in New York.)
I also liked that it didn't explicitly say how it decides when to play a note.
All the subway routes are normalized to 15 seconds long from beginning to end. The app then plays all 15 second routes together, playing the instrument assigned to the route when there's a train there.
Neat commentary on the instruments that were assigned to the route when you mouse over it.
Can someone explain what triggers a note? I don't understand from the explanation on the site. Is it whenever a train on the line crosses a predetermined geo-location?
What do the technical details look like behind this to get the data?
Very neat. This is an example of digital art that I’d love to see exist in physical form somehow. I suppose it could get rather noisy at a museum but I love the intersection of mass transit & music.
I like that Americans, and particularly New Yorkers, are so self-centered that they don't even bother to specify which city's train lines are on the map (New York is only in the HN title). Of course, a Ukrainian frontend developer will be familiar enough with the city's contours to instantly recognize the shape of Manhattan from the train lines.
I’d love to see this as an art installation in a subway station, please consider chatting with MTA Arts & Design, maybe they could hook you up with the right people to make this a fixture somewhere: https://www.mta.info/agency/arts-design
Wow! I really like this! I have been practicing with sounds in a similar context: I have made the GitHub contribution graph as an space visualization and a music generated based on the graph [1]
Hey, this is amazing. I've been building another musical toy that I'm terming Euclidean Pulses, but I haven't been able to find a good library for making sounds. What did you use?
It's interesting that a lot people like this but dislike AI-generated music. The music itself here is completely random to us, yet I can't see how AI-generated music can be worse than random.
The idea is novel/fun/cool, but the notes ARE random as far as we can tell. So if you're against AI music, you just like the idea but don't care about the music or... something else I can't imagine.
I think we can all come up with a bunch of original "hey, if we turn this random pattern of X into music, it would be interesting". But I don't see the point of actually doing it since the result is obviously going to be random uninteresting notes. If I convert my keypresses on my keyboard over the past year or whether my dog licks itself or barks or runs into music, it would still be random crap. The idea of the article is the only thing that made me go "huh" for a few moments. Clicking around and seeing the execution and hearing the music was definitely "meh".
The music all by itself is not particularly enjoyable here. What's great is the concept, execution, and the way data from an unlikely source is directly audible in the music. What defines art will always be fuzzy, but this particular work is a good example of art I can appreciate: presenting known things in an unusual way, playing with perception to create new connections between remote concepts, and sometimes providing a stepping stone to, as you say, enlightenment.
This is not random in the slightest. Each instrument was carefully chosen based on characteristics of the line. The notes were placed along the line by a human. Each step of the way involved a human making choices. The underlying driver… the trains locations are on a schedule.
There are variations as trains run fast or slow or not at all. Even those events are results of causes.
It might not be repeatable or predictable but it is not random.
Also, an artist made this. I can appreciate the design and flair of another human. AI is soulless. And there was a nothing to celebrate. No one to clap on the back and say “good job”. No one to identify with and say “people are really neat.”
To me the question is more about why you'd do something. I'm sure that there's a lot of AI generated music I might enjoy, but I'm turned off by why it exists.
AI generated music isn't out there as some experiment by some artists, trying to make sense of something, make a statement, or just for the sheer enjoyment of creating something. It's there because of money. I know, there are exceptions, and I'm fine with those.
The AI music I'm against is the type that's made by the likes of Spotify, because they don't want to pay artists. It's music that only exists because Spotify would like to make more money. That motivation, to me, corrupts the product. AI music isn't created for the sake of creating music, it's created as a means towards a goal, money. I don't think that the management at Spotify particularly cares about music, it's just a means on the path to money. If they could be more profitable selling something else they would.
You can argue that a lot of bands solely exists because a record labels wanted to create a profit generating device, or that some artists are artificially created and wouldn't survive without auto-tune and a media machine pushing their music. I think that's bad as well, but often many of these artists do have some level of talent and actually do care about music.
I don’t actually care for jazz. But I like this for the concept. I listed to this longer than any other jazz I had the option to turn off. Just to explore the results and learn about the different lines. Music, art really, includes far more than the notes, or finished product.
Bolero is an amazing piece of music. Ravel’s brain was suffering from a degenerative disease at the time. We would not have Bolero without his disease. That fact to me turns the piece of music into a meditation on what his mind may have been like. What it might have been to be Ravel.
Why do you think people dislike AI-generated content?
It's not because AI-generated music inherently sucks. It's generally C-grade professional music. It's just not novel or especially interesting, and the low barrier to entry means there's a ton of slop in the space.
A lot of people have always wanted to make music, never made it past the barrier of "music is hard," and therefore have no clue as to what makes truly good music. And now that they have AI, they think they can just skip all the boring parts and make great songs.
And while they can skip a lot of steps in the creative process — those skipped steps also help musicians develop their artistic taste and judgment.
And just because these AI "creators" can't tell the difference, they assume others can't either. And then they get mad when critics recognize their uninspired, derivative slop for what it is.
That's not limited to music, either. You see it in coding, graphic design, writing, and pretty much any other LLM-assisted content generation. Maybe it'll change one day as models get better. Maybe not.
This project is original, stylish, technically clever, aesthetically pleasing, and well-crafted. There's a level of polish and intention behind it, and people here recognize that.
I think it has to be with expectations. Out of random music we don't expect much, so any result that is nice is good enough. For AI we are promised it's "just as good" but we get generic, soulless music that bring nothing new to the table.
Yeah, it's better than a lot of people, but it doesn't deliver the "just as good" part. On top of that you get that now anyone can promp a song and have a deluge of grey, tasteless elevator music.
If I could offer one usability suggestion: darken the text displayed at the bottom when a specific route is selected. Currently it's much too light for the white background. I couldn't tell you the exact contrast ratio but I'm certain it doesn't meet accessibility guidelines.
The sound doesn't work for me, but I love the description of the G. It does have a cult following: and just like a cult you're tricked into loving it despite its many flaws, like the one hour wait at night or sprinting to the middle of the platform.
75 comments
I would love to see a Japanese version someday. Tokyo’s train network is one of the most complex in the world, and I imagine the music it would make would be extraordinary. Thank you for creating something so thoughtful.
But then, jazz is sometimes spoken of as expressing the rhythms, sounds, and emotions of the modern city.
I also liked that it didn't explicitly say how it decides when to play a note.
All the subway routes are normalized to 15 seconds long from beginning to end. The app then plays all 15 second routes together, playing the instrument assigned to the route when there's a train there.
Neat commentary on the instruments that were assigned to the route when you mouse over it.
What do the technical details look like behind this to get the data?
[1] https://joaonetto.me/projects/music
(the mute control in the lower right is always enabled)
Reminds me of SnakeJazz
The idea is novel/fun/cool, but the notes ARE random as far as we can tell. So if you're against AI music, you just like the idea but don't care about the music or... something else I can't imagine.
I think we can all come up with a bunch of original "hey, if we turn this random pattern of X into music, it would be interesting". But I don't see the point of actually doing it since the result is obviously going to be random uninteresting notes. If I convert my keypresses on my keyboard over the past year or whether my dog licks itself or barks or runs into music, it would still be random crap. The idea of the article is the only thing that made me go "huh" for a few moments. Clicking around and seeing the execution and hearing the music was definitely "meh".
Enlighten me, please.
There are variations as trains run fast or slow or not at all. Even those events are results of causes.
It might not be repeatable or predictable but it is not random.
Also, an artist made this. I can appreciate the design and flair of another human. AI is soulless. And there was a nothing to celebrate. No one to clap on the back and say “good job”. No one to identify with and say “people are really neat.”
AI generated music isn't out there as some experiment by some artists, trying to make sense of something, make a statement, or just for the sheer enjoyment of creating something. It's there because of money. I know, there are exceptions, and I'm fine with those.
The AI music I'm against is the type that's made by the likes of Spotify, because they don't want to pay artists. It's music that only exists because Spotify would like to make more money. That motivation, to me, corrupts the product. AI music isn't created for the sake of creating music, it's created as a means towards a goal, money. I don't think that the management at Spotify particularly cares about music, it's just a means on the path to money. If they could be more profitable selling something else they would.
You can argue that a lot of bands solely exists because a record labels wanted to create a profit generating device, or that some artists are artificially created and wouldn't survive without auto-tune and a media machine pushing their music. I think that's bad as well, but often many of these artists do have some level of talent and actually do care about music.
Bolero is an amazing piece of music. Ravel’s brain was suffering from a degenerative disease at the time. We would not have Bolero without his disease. That fact to me turns the piece of music into a meditation on what his mind may have been like. What it might have been to be Ravel.
It's not because AI-generated music inherently sucks. It's generally C-grade professional music. It's just not novel or especially interesting, and the low barrier to entry means there's a ton of slop in the space.
A lot of people have always wanted to make music, never made it past the barrier of "music is hard," and therefore have no clue as to what makes truly good music. And now that they have AI, they think they can just skip all the boring parts and make great songs.
And while they can skip a lot of steps in the creative process — those skipped steps also help musicians develop their artistic taste and judgment.
And just because these AI "creators" can't tell the difference, they assume others can't either. And then they get mad when critics recognize their uninspired, derivative slop for what it is.
That's not limited to music, either. You see it in coding, graphic design, writing, and pretty much any other LLM-assisted content generation. Maybe it'll change one day as models get better. Maybe not.
This project is original, stylish, technically clever, aesthetically pleasing, and well-crafted. There's a level of polish and intention behind it, and people here recognize that.
Yeah, it's better than a lot of people, but it doesn't deliver the "just as good" part. On top of that you get that now anyone can promp a song and have a deluge of grey, tasteless elevator music.
I sent my friend an AI generated album and he loved it, until I mentioned it was AI generated, and then he went a bit pale.