So i tried to post a small program that directly creates a wav file, but it was
blocked.

The idea is that since any sound, track, music , album is just a sequence of numbers, why not just create that sequence through a program ?

So my program created a sequence of samples for a PCM encoded wav file
and writes them to a wav file which can be heard with windows media player.

Granted the possible combinations are huge between the the design of any number generating program and the sequence, but I think many never heard before sounds and "music" or "songs" can be created this way.

Has it ever been tried before ?

Searching the internet doesn't give many results.


Thanks if anyone knows if it has been experimented.

I think the file extensions you are allowed to post are listed: wavs mp3, jpegs, pngs. etc. If it is not on the list, the forum won't let you attach it. In any case, "a small program that directly creates" something strikes me as really similar to a virus, so maybe that's why it was blocked!

well if this sequence of numbers is randomly generated, I think it is going to sound horrible. The numbers in a wav file are supposed to represent the position of the sound wave at 'X' point in time. To appear as a coherent sound, the position of the sound wave at the next point in time would normally be represented by a number that is slightly higher or slightly lower than that original number. Rising and falling gradually enough to be heard as a sound by human beings. Not jumping up and down randomly across the zero line every .02 ms.

yeah, but if you have heard one set of random hash you have pretty much heard them all. Wouldn't random numbers sound like the static between stations? I would assume that only combinations that rise and fall in some coherent way will 'sound' like anything.
I thought you just did it? Didn't you just make this program? You tell us. How does it sound? I am sure the forum software would allow you to post the resulting output -- as a wav file or mp3, even though it blocks the executable program itself.

If you do, I am going to turn my volume way down before clicking on it.

Interesting to use in noise music, contemporary classical music, avant gard, experimental etc.

I am using sin and cos functions and others. I have many wav files that generate sounds and sound effects.

So the combinations are never ending between the code you can invent etc.

some python code snippets,

dur.append(a)
note.append(n)
for x in range(200000):
b.append(int(32760 * (math.sin(1000 * x) + math.sin(100 * x))))

y = int('0'+note[i])
d = int('0'+dur[i])

d=int(d*y/t)
y=int(y/t)

for l in range(d):
k2=b[l]

for x in range(y):
new.append(k2)

So the note is created by stretching the wave (range(y))

and length of note is done by number of cycles (range(d))

And so you can experiment by changing anything you want, etc.

So Ai is overkill, and in fact the best Ai music is the ones that make mistakes and mess up the sound!

A short example, but I have many more etc.


So the idea is any possible album or track can be made like this, in theory, but this is where something interesting can come from, not AI imitating songs we heard hundreds of times.

Yes, though computer generated music has been around considerably longer than WAV files. We were using a DEC PDP 11 and an IBM 360 way back in the mid 70’s.

Over the years, I’ve used assembly, (pre-Microsoft) basic, Fortran, C++, MATLAB, etc.

Maybe search GitHub for music synthesizers? Or grab an LLM and press it for historical context?

Also, maybe setup a project and website on GitHub—then post a link here.


Keep pressing forward,

Ray H.

OK that is far more 'coherent; than I expected. I guess I expected something like white noise. This has a structure that does not seem completely random.

I would like to hear some more examples. Especially ones that show of the variety of sounds that this can produce.

Ai is the wrong way to do it. It creates similar results from
previous results.

The tecnique of directly writing the samples
of a wav file creates new and interesting results, not tied to what humans think are good: it is completely free to create anything.

with 30 to 50 lines of code you cancreate novel sounds and "songs", but the combinations are endless.

If you want to hear "familiar" things than humans have already written a million tracks from 1965 (when modern music really started) to today (in fact many tracks are hard to pinpoint down what era they are, if 1975 or 1995 or 2015 etc).

Ai seems to create thousand of new tracks etc.

I think that by experimenting with my method of writing the samples directly you can create a never ending source of new "songs" (40,000 samples a second between -30000 and + 30000 amplitude etc.).

Here is another example attached.

So you can see the perl file here:

https://github.com/eightm71/wav-files/blob/main/pltxt

this can be executed as c: \ perl pltxt.pl 333

and creates a wav file

Well, no offense but that sounds an awful lot like the first one to me. Beeps and boops. Like tuning a shortwave radio really fast. No one would be surprised to learn that both files were produced from the same "engine". The second one sounds like something that happened five seconds after the first one cut off.

I was kind of hoping to hear something completely different. I don't think I personally would have an application for this.

I could have 5 different colors of jellybeans - red, yellow, green, orange, black - and I could make "endless" combinations of those jellybeans. Two greens, two reds and one yellow. But these combinations will always only taste like jellybeans.

None of those combinations are going to taste like pepperoni - or broccoli.
.

As a software dev, one of my all time favorite books is ‘Analytic Combinatorics’ by Philippe Flajolet and Robert Sedgewick. [1]


The text was more focused on performance than tastes, though.

Ray H.

@ eightm71 —

Chew on @ joeq ‘s feedback—even if your goal is beep-boop avant-garde.

Music generally requires that we build some kind of kind of vocabulary. And then build meaningful structure on the vocabulary.

It’s not yet sufficiently apparent in your artifacts what those structures and vocabulary are to be.


[1] https://ac.cs.princeton.edu/home/

Yes, i understand your point of view.

But so much music doesn't have much variety anyways

and this is an attempt to create some directly.

if you take a wav file and stretch it through another program I
have you can get a somewhat different song (or sound effect, or free "noise rock" track).

here attached are the results.

https://github.com/eightm71/wav-file...ain/pltxt2.txt

The x300 file is just a wav file header, if you look at the code.

The only problem is exploring many file to find the snippets that you like and

can extract with a DAW or whatever and create another song.

If you like programming, you can try it with perl and experiment,

I'm happy with the idea that any sound someone labels music is music.

So what do I make of these? The first is a little bit like Stockhausen, the others sound most like a dial up modem or something lost on Autechre's hard drive.

So at the moment this is pretty Avant Garde. It sounds like one stream of stuff, just monophony and I'm not sure that I would call them songs. It would be interesting to see if the technique could produce polyphony and allow for structure and rhythm. As it is, I'm struggling to see an application for it, Refine and use as a sample creator, integrate something to help sift the good from the bad, stitck a friendly front end on it?

It's interesting to see you experimenting OP and thank you for sharing.

Well this started as an image generation program.

It was developed after a few years, running from DOS
command line in perl. It was edited and the resulting image was seen in vim (the vi editor for windows DOS).

The programs are under imgg.txt and the resulting image
can be seen by opening imgg with vim.

Only with DOS and vim you can see a good image.

The image seems like music for the eyes, so the idea came to try to make a music program-

Anyways the image could be the notation or chart for some kind of music, by associating it with instruments or
musical ideas, but I still don't have any program that does that (i,ll think about it) or if some of you like programming try to invent something.


the github repository:


https://github.com/eightm71/wav-files

"notation" to me suggests MIDI. If your program generated MIDI notes, of different pitches and durations and velocities they could be restricted to the 12-tone scale of a piano - or not - but at least you could have the option to generate something more 'accessible' to the average listener.

To go back to the sounds, it seems me some kind of "smoothing" might be useful. Instead of crackling up and down like a fire at a fireworks factory, have it produce more continuous or 'slowly evolving' sounds. I have no idea what that would require in terms of programming, but I bet more people would find it useful.