Show HN: MTXT – Music Text Format (opens in new tab)

(github.com)

124 pointsdaninet3mo ago39 comments

39 comments

How does this compare to standard ABC? More capable, presumably, but a comparison would be useful.

https://en.wikipedia.org/wiki/ABC_notation https://abcnotation.com/

ABC notation is more oriented towards traditional sheet music, with regular note lengths, standard Western tuning and a simple, readable syntax. It isn't meant for playing back music that sounds good to the ear. It's hard to catch the nuances of a real human performance with it, but it works well as a lead sheet for musicians. Its expressive marking are relatively limited and interpreted subjectively.

MTXT focuses on editable recordings of live performances, preserving all of those tiny irregularities that make the music human. It can represent arbitrary timings, subtle expressive variations and even arbitrary tuning systems. MTXT can also capture transitions like crescendos and accelerandos exactly as they happened.

dghf3mo ago

Similar things:

* Perl MIDI::Score -- https://metacpan.org/pod/MIDI::Score

* Csound standard numeric scores -- https://csound.com/docs/manual/ScoreTop.html

* CsBeats (alternative score language for Csound) -- https://csound.com/docs/manual/CsBeats.html

bonzini3mo ago

Lilypond, too. Though it needs a full scheme interpreter to evaluate macros (provided by both the system and the user), it can emit midi files.

ramses03mo ago

Lilypond isn't well-known enough!

https://en.wikipedia.org/wiki/LilyPond#Integration_into_Medi...

https://www.mutopiaproject.org

https://lilypond.org/text-input.html

     \relative c' { 
        \key d
        \major
        fis4 fis g a
        a g fis e
        d d e fis
        fis4. e8 e2
     }

...but why is it so complicated? A novice interpretation of "music" is "a bunch of notes!" ... my amateur interpretation of "music" is "layers of notes".

You can either spam 100 notes in a row, or you effectively end up with:

    melody   = [ a, b, [c+d], e, ... ]
    bassline = [ b, _, b,     _, ... ]
    music = melody + bassline
    score = [
       "a bunch of helper text",
       + melody,
       + bassline,
       + page_size, etc...
    ]

...so Lilypond basically made "Tex4Music", and the format serves a few dual purposes:

Engraving! Basically "typesetting" the music for human eyeballs (ie: `*.ly` => `*.pdf`).

Rendering! Basically "playing" the music for human ears (ie: `*.ly` => `*.mid`)

Librarification! Basically, if your music format has "variables" and "for-loops", you can end up with an end score that's something like: `song = [ intro + chorus + bridge + chorus + outro ]`, and then not have to chase down and modify all the places you use `chorus` when you modify it. (See this answer for more precision: https://music.stackexchange.com/a/130894 )

...now imagine doing all of the above for multiple instruments and parceling out `guitar.pdf`, `bass.pdf`, `drums.pdf` and `whole-song.pdf`

TL;DR: Music is haaard, and a lot closer to programming than you think!

1 more reply

vessenes3mo ago

Cool. My one concern with this is that it has no horizontally scannable note/chord mode. It’s super common for humans to read a sequence of notes left to right, or write it that way, but it’s also just more efficient in terms of scanning / reading.

Can I suggest a guarded mode that specifies how far apart each given note/chord is by the count, e.g.

  #1.0:verse1 
  Am - C - G - E - F F F F
  #

You could then repeat this or overlay a melody line like

  #0.25:melody1
  C4 - C4 - C4 D4 C4 - D4 - D4 - D4 E4 D4 -
  #

Etc. I think this would be easier to parse and produce for an LLM, and it’s would compile back to the original spec easily as well.

daninetOP3mo ago

I considered it but decided against it in the first version, because specifying note durations is too tricky. It was more important to get the .mid -> MTXT conversion and live-performance recording working, where notes usually have irregular note lengths. Representations like "C4 0.333 D4 0.333 E4 0.25" feel too hard to read.

matheusmoreira3mo ago

To me it seems like files could get hard to understand if events that happen simultaneously aren't horizontally lined up like this:

  2.0 voice1 | voice2 | ...

Like a text version of old school tracker interfaces:

https://youtu.be/eclMFa0mD1c

  POS | TRACK #1 | TRACK #2 | ...

HelloNurse3mo ago

Aren't the notes adjacent enough on consecutive lines?

  2.0 note Cmaj7 ch=1 
  2.0 note D ch=1 
  2.0 note C dur=0.15 ch=2
  2.1 note C ch=2
  2.1 note Cmaj ch=1

matheusmoreira3mo ago

Imagine there are four, eight, maybe dozens of voices being mixed together in a track. Could get unwieldly.

1 more reply

Grom_PE3mo ago

This made me remember old set of tools called mtx2midi and midi2mtx, I used them to edit some midi files while making sure I'm not introducing any unwanted changes. While roundtrip output was not binary identical, it still sounded the same.

Looks like MTXT tool here does not quite work for this use case, the result of the roundtrip of a midi I tried has a segment folded over, making two separate segments play at the same time while the total duration got shorter.

https://files.catbox.moe/5q44q0.zip (buggy output starts at 42 seconds)

daninetOP3mo ago

Thank you, I will have a look. I consider it important to have the round trip conversion working seamlessly.

I created an issue here: https://github.com/Daninet/mtxt/issues/1

cestith3mo ago

It reminded me of ABC and the tools abc2midi and midi2abc.

chaosprint3mo ago

Some simple thoughts:

I feel that one challenge of programming languages is how to remember these rules, formats, and keywords. Even if you're using familiar formats like YAML or JSON, how do you match keywords?

When developing Glicol (http://glicol.org/), I found that if it's based on an audio graph, all node inputs and outputs are all signals, which at least reduces the matching problems. The remaining challenge is ensuring that reference documentation is available at the minimal cost.

rock_artist3mo ago

Hey, the idea is nice, It would be great to know what pushed you to start this format.

Also, any apps that uses it would benefit from being add to the repo assuring usability in addition to readibility.

daninetOP3mo ago

My initial goal was to fix some mistakes in the MIDI files I recorded from my keyboard. I was also interested in making dynamic tempo and expression changes without dealing with complicated DAW GUIs.

Now I'm working on a synth that uses MTXT as its first-class recording format, and it's also pushing me to fine-tune a language model on it.

1313ed013mo ago

I like the idea overall. Looks like something that would be fun to combine with music programming languages (SuperCollider/Of etc).

Not so sure how human-friendly the fractional beats are? Is that something that people more into music than I am are comfortable with? I would have expected something like MIDIs "24 ticks per quarter note" instead. And a format like bar.beat.tick. Maybe just because that is what I am used to.

daninetOP3mo ago

The library has MIT license, I would be more than happy to see people use it in different synths.

I'm planning to add support for math formulas in beat numbers, something like: "15+/3+/4" = 15.58333

soperj3mo ago

> "15+/3+/4"

Can you explain how to read that? 15 plus divided by 3 plus divided by 4?

1 more reply

bonzini3mo ago

It should be fine, but fractions (or both fractions and decimals) would be preferable in order to express triplets (3 over 2, effectively a duration of 0.3333...)

intrasight3mo ago

I think that for completeness it needs looping and conditional constructs

jasonjmcghee3mo ago

This would lend itself well to a live-coding/live-music experience.

I played around with a similar idea on my own (very simple / poor) text music environment:

https://github.com/jasonjmcghee/vscode-extension-playground?...

in the middle of making an extension to allow making vs code extensions live because I wanted a faster development feedback loop.

usrusr3mo ago

Count me in as another one with a longstanding mostly dream project aiming for human enjoyable notation grammar.

For me it was coming from tracker notation (buzz), where i was wildly underwhelmed by all that whitespace for timing (well, empty cells for timing) and the lack of parameterizable macros. A seriously underexplored field, perhaps because almost everybody who ever started got pulled in by the lure of textually defined synthesis.

chrisjj3mo ago

Solved 40yrs ago. AMPLE.

http://www.colinfraser.com/m5000/ample-nucleus-pg.pdf

For macros, see:

https://www.retro-kit.co.uk/user/custom/Acorn/3rdParty/Hybri...

formula13mo ago

This is pretty neat

I'm wondering if it can be used alongside strudal https://strudel.cc/ Either mtxt => strudal or strudal => mtxt

Heres strudal in action https://www.youtube.com/shorts/YFQm8Hk73ug

xrd3mo ago

I've been spending the last week casually looking at strudel.cc.

They have a notation that looks similar (basically a JavaScript port of the Haskell version).

I like this, but I'm curious why I would want to use this over strudel. Strudel blends the language with a js runtime and that's really powerful and fun.

lokar3mo ago

I have been using:

https://www.vexflow.com/

Which has a text format, and typesets it for you nicely.

throw73mo ago

It makes no sense to design for llm's. Do what makes sense for the reader and forget that llm's exist at all.

amingilani3mo ago

What prompted this and why does it not?

badlibrarian3mo ago

It's not the 19th Century. You don't need to punch holes in cards to help the machine "think" any more.

1 more reply

yaoke2593mo ago

pretty cool!

jan_Sate3mo ago

Obligatory xkcd: https://xkcd.com/927/

giladvdn3mo ago

Probably stating the obvious here, but this would be a good way for an LLM to attempt to write or modify music.

j / k navigate · click thread line to collapse

39 comments

gilrain3mo ago

How does this compare to standard ABC? More capable, presumably, but a comparison would be useful.

https://en.wikipedia.org/wiki/ABC_notation https://abcnotation.com/

daninetOP3mo ago

dghf3mo ago

Similar things:

* Perl MIDI::Score -- https://metacpan.org/pod/MIDI::Score

* Csound standard numeric scores -- https://csound.com/docs/manual/ScoreTop.html

* CsBeats (alternative score language for Csound) -- https://csound.com/docs/manual/CsBeats.html

bonzini3mo ago

Lilypond, too. Though it needs a full scheme interpreter to evaluate macros (provided by both the system and the user), it can emit midi files.

ramses03mo ago

Lilypond isn't well-known enough!

https://en.wikipedia.org/wiki/LilyPond#Integration_into_Medi...

https://www.mutopiaproject.org

https://lilypond.org/text-input.html

     \relative c' { 
        \key d
        \major
        fis4 fis g a
        a g fis e
        d d e fis
        fis4. e8 e2
     }

...but why is it so complicated? A novice interpretation of "music" is "a bunch of notes!" ... my amateur interpretation of "music" is "layers of notes".

You can either spam 100 notes in a row, or you effectively end up with:

    melody   = [ a, b, [c+d], e, ... ]
    bassline = [ b, _, b,     _, ... ]
    music = melody + bassline
    score = [
       "a bunch of helper text",
       + melody,
       + bassline,
       + page_size, etc...
    ]

...so Lilypond basically made "Tex4Music", and the format serves a few dual purposes:

Engraving! Basically "typesetting" the music for human eyeballs (ie: `*.ly` => `*.pdf`).

Rendering! Basically "playing" the music for human ears (ie: `*.ly` => `*.mid`)

...now imagine doing all of the above for multiple instruments and parceling out `guitar.pdf`, `bass.pdf`, `drums.pdf` and `whole-song.pdf`

TL;DR: Music is haaard, and a lot closer to programming than you think!

1 more reply

vessenes3mo ago

Can I suggest a guarded mode that specifies how far apart each given note/chord is by the count, e.g.

  #1.0:verse1 
  Am - C - G - E - F F F F
  #

You could then repeat this or overlay a melody line like

  #0.25:melody1
  C4 - C4 - C4 D4 C4 - D4 - D4 - D4 E4 D4 -
  #

Etc. I think this would be easier to parse and produce for an LLM, and it’s would compile back to the original spec easily as well.

daninetOP3mo ago

matheusmoreira3mo ago

To me it seems like files could get hard to understand if events that happen simultaneously aren't horizontally lined up like this:

  2.0 voice1 | voice2 | ...

Like a text version of old school tracker interfaces:

https://youtu.be/eclMFa0mD1c

  POS | TRACK #1 | TRACK #2 | ...

HelloNurse3mo ago

Aren't the notes adjacent enough on consecutive lines?

  2.0 note Cmaj7 ch=1 
  2.0 note D ch=1 
  2.0 note C dur=0.15 ch=2
  2.1 note C ch=2
  2.1 note Cmaj ch=1

matheusmoreira3mo ago

Imagine there are four, eight, maybe dozens of voices being mixed together in a track. Could get unwieldly.

1 more reply

Grom_PE3mo ago

https://files.catbox.moe/5q44q0.zip (buggy output starts at 42 seconds)

daninetOP3mo ago

Thank you, I will have a look. I consider it important to have the round trip conversion working seamlessly.

I created an issue here: https://github.com/Daninet/mtxt/issues/1

cestith3mo ago

It reminded me of ABC and the tools abc2midi and midi2abc.

chaosprint3mo ago

Some simple thoughts:

I feel that one challenge of programming languages is how to remember these rules, formats, and keywords. Even if you're using familiar formats like YAML or JSON, how do you match keywords?

rock_artist3mo ago

Hey, the idea is nice, It would be great to know what pushed you to start this format.

Also, any apps that uses it would benefit from being add to the repo assuring usability in addition to readibility.

daninetOP3mo ago

My initial goal was to fix some mistakes in the MIDI files I recorded from my keyboard. I was also interested in making dynamic tempo and expression changes without dealing with complicated DAW GUIs.

Now I'm working on a synth that uses MTXT as its first-class recording format, and it's also pushing me to fine-tune a language model on it.

1313ed013mo ago

I like the idea overall. Looks like something that would be fun to combine with music programming languages (SuperCollider/Of etc).

daninetOP3mo ago

The library has MIT license, I would be more than happy to see people use it in different synths.

I'm planning to add support for math formulas in beat numbers, something like: "15+/3+/4" = 15.58333

soperj3mo ago

> "15+/3+/4"

Can you explain how to read that? 15 plus divided by 3 plus divided by 4?

1 more reply

bonzini3mo ago

It should be fine, but fractions (or both fractions and decimals) would be preferable in order to express triplets (3 over 2, effectively a duration of 0.3333...)

intrasight3mo ago

I think that for completeness it needs looping and conditional constructs

jasonjmcghee3mo ago

This would lend itself well to a live-coding/live-music experience.

I played around with a similar idea on my own (very simple / poor) text music environment:

https://github.com/jasonjmcghee/vscode-extension-playground?...

in the middle of making an extension to allow making vs code extensions live because I wanted a faster development feedback loop.

usrusr3mo ago

Count me in as another one with a longstanding mostly dream project aiming for human enjoyable notation grammar.

chrisjj3mo ago

Solved 40yrs ago. AMPLE.

http://www.colinfraser.com/m5000/ample-nucleus-pg.pdf

For macros, see:

https://www.retro-kit.co.uk/user/custom/Acorn/3rdParty/Hybri...

formula13mo ago

This is pretty neat

I'm wondering if it can be used alongside strudal https://strudel.cc/ Either mtxt => strudal or strudal => mtxt

Heres strudal in action https://www.youtube.com/shorts/YFQm8Hk73ug

xrd3mo ago

I've been spending the last week casually looking at strudel.cc.

They have a notation that looks similar (basically a JavaScript port of the Haskell version).

I like this, but I'm curious why I would want to use this over strudel. Strudel blends the language with a js runtime and that's really powerful and fun.

lokar3mo ago

I have been using:

https://www.vexflow.com/

Which has a text format, and typesets it for you nicely.

throw73mo ago

It makes no sense to design for llm's. Do what makes sense for the reader and forget that llm's exist at all.

amingilani3mo ago

What prompted this and why does it not?

badlibrarian3mo ago

It's not the 19th Century. You don't need to punch holes in cards to help the machine "think" any more.

1 more reply

yaoke2593mo ago