pouët.net

I am preparing a 4k intro, and i need some help on music. I want something ambient, no drums ntoup-ntoup-ntoup etc at all.
1. What is the target size of music in 4k intros generally.?
2. What is more size efficient, midi or custom synth?
3.Is out there a tool to experiment with procedurally generated sounds?

Sorry if i sound naive, i know nothing about music production.
Thanx.

Just had to say that you have good taste, sir.

What serpent said. I've been experimenting with 4klang for ambient soundtrack for a 4k intro, and it works quite well.

keywords: echo and reverb if you want ambient sounds.

Also, midi would of course be taking less space, but you would still need some post-processing in order to achieve nice "ambient" sounds in my opinion.

I'm pretty sure that you could be able to create nice ambient sounds with 4klang.

I havn't been doing music for 4k intros with 4klang, but usually we spend around 1000-1500 bytes on synth+music data in our intros.
Possibly ambient sounds should take up less data space, since there is less information to store regarding notes etc.
Of course this always depends on the complexity, but there should be some data to save by not having a bassline, beats etc.

I think music should be allowed to take up 25% of an intro size.

MIDI is more size efficient, but good luck making it sound right.

All (okay, most) music synths are procedural, actually. I usually ask this question and people are often perplexed by it, but do you YOURSELF make music?

using 4klang, it's more like the music takes 2000-2500 bytes with the synth from my experience.

4klang used right should end up at 1400 - 1500 compressed bytes including a song. compiling an exe with pe header and crinkler stub will take this to about 2000 and up bytes, yes.
rough numbers anyway :)

I think that using 4klang for simple ambient would be an overkill. You could do some very simple synth (few waveforms, echo/delay, simple reverb) yourself for your specific case in far less bytes. For example, a custom-made synth in my 1k for linux does very simple AR-envelope, fakes reverb by sustain, echoes output (delay) and takes about 160 gzip-compressed bytes with sequence included.

Before you go ahead and spend a lot of bytes on something simple, try some of the techniques from this with some aggresive low pass filters and heavy reverbs/echos.

midi or custom synth?
what does that mean?

midi is just an interface, the synth is stand alone.
so you need both! (unless i am missing something here).

good luck anyway! and what Serpent said!

If you want to go the experimental route, you can generate good atmospheric sounds using very large (inverse) Discrete Fourier Transforms. The DFT routine is very compact (but slower than the FFT).

Look here for code inspiration and here and here for audio.

You probably won't get that far (as the examples) without at least an ADSR-envelope for amplitude and definately some kind of filter. And for 4k's I guess everything counts ;)

In the wonderful land of wavetable-oscillators, each operation for rendering the wavetables will consume space as well (and very likely code to handle a bit of precalc as well) Not sure on the last part, but the more complex operations you apply for generating wavetables, the more likely you're going to want to precalc it.

I can personally see music using this algorithm fitting into 4k and still leaving some space for effects too. Maybe I'm wrong, but I hope someone will prove me right (I don't have the skills as of now to make a demo).

it's possible to use 4klang... with like, one instrument that you multiply the number of notes you'll want per chords * 2 as you don't want the note to stop when an other come in right?

So, triad it will be. 3*2 -> already 6.
Gives you +2 for a nice lead.

And you can also make a very nice "send channel" if you fiddle with some combo like this: reverb->disto->delay->verb (insert a flanger somewhere xD)

What about combining formulars from the "Experimental music from very short C programs " thread. You might have one "channel" with a rhythmic formular, one channel with a drone sound and one that generates some blips and bleeps. You could combine that with simple effects.

Or a synth with a wavetable generated by a formular...

I'm no programmer, so I don't know if that actually sums up to be more bytes than having a synth and simple note/pattern data. It might sound a little different and more unexpected though. :)

Very interesting what you guys accomplish here!

I strongly suggest you to do your own synth. Mostly because it's a fun thing to code and that your needs are very specific, so you'll achieve better in less bytes.

Audio size: in the 4Ks I've done the audio ended up pretty big - 2000b w/ data - , mostly because I was trying to make it too versatile.

Like Lord Graga and varjohukka said, short semi-random algo with heavy effect could go a long way while staying super small. Only problem here is if you're just the coder and your musician wants all the freedom in the world, or sometimes worst: already has the track done and you need to "port" it.