You can’t get exact synchronisation of MP3s because the encoder/decoder delay is not defined. Most other formats (including WAV, FLAC and OGG Vorbis) don’t have this problem.
Probably better to use hardware that is designed for the job. Although Audacity can only play as 2 channel stereo, multi-channel sound cards are readily available, including inexpensive sound cards designed for “surround sound”. Audacity is able to produce multi-channel audio files even though it can only play 2 channel stereo (see: Advanced Mixing Options - Audacity Manual)
Most DAW applications and some media players (including VLC) can play multiple audio channels if suitable hardware is present and working correctly.
Such software exists, but lacks mainstream popularity because creating music in such a detailed way is incredibly arduous.
If you want to be totally without restriction, then all things must be allowed, and it is the matter of defining the specific thing that you want that is arduous.
For example, It is common for synthesizers to offer a selection of simple cyclical waveforms, such as Sine, Triangle, Square, Sawtooth. All cyclical waveforms may be defined in terms of a series of pure (sine) waves, their frequency, amplitude and phase. For example, a square wave can be defined as a series of odd harmonics of decreasing amplitude:
So let’s take away the restriction and say that we allow ‘any’ series of sine waves that have frequency, amplitude and phase related (or not) by any rules that the composer wants.
That takes us from a simple 4 way choice betwee Sine, Triangle, Square, Sawtooth, to defining an arbitrary algorithm for generating any number of sine waves of any amplitude and any phase (and there are of course an infinite number of different waveforms that could be generated).
That is much more “arduous” than simply pressing a key on an electronic keyboard, but much less restricted in that we have precisely defined the harmonic content that we want, rather than just selecting a waveform that the keyboard manufacturer thought we should have.
One way to approach the problem is to use a graphical grid. Let’s suppose we want to create a few sounds that have integer multiple frequencies for overtones. Few notes will have more than 32 partials, but let’s take the severe case of a base instrument having seven octaves of sine wave components (partials). That would be 128 frequencies. The horizontal axis of the graphical grid would have the numbers 1 through 128, or 1 through 256 if you like. As you move the pointing device from left to right or right to left, the cursor jumps from one number to its neighbor without giving you options between. As you move the cursor vertically, every value is an option, possibly accompanied by a multi-digit number window. As you scan through these vertical places, the last vertical value touched before moving onto another horizontal position is kept as the value for that partial (the numbered on the horizontal axis).
Sound feedback could be sampled at any time for the waveform as chosen so far. Experience would quickly reveal that some sorts of paths across the grid make no sense, and further experience would reveal that some sorts of paths seem inappropriate for the music you are writing. In eight hours, you might seriously consider 40 sounds, but after a month, maybe you will like 15 or maybe just 4. This is not tedium. This is freedom.
If you really want a fast but less exacting method, just draw a waveform where time is the horizontal axis and intensity is the vertical axis and listen to it.
There are, of course, many short-cut features that could be on the menu. A clarinet has only odd numbered partials. Clicking off or attenuating even numbered partials should not require a lot of up and down drawing. Also, there are curves that could be applied and superimposed. If you wanted octaves of the second partial to be louder than the octaves of the third or fifth partial all the way up, this should not necessarily be left to free hand drawing. I’m sure you can image many conveniences or this sort.
There is another important issue regarding the computer aided design of sounds. When the user inputs a drawn waveform, that waveform might have an instantaneous change from one vertical position to another, and it might sound good. A saw tooth wave is an example. Speakers cannot change position instantaneously. If all of the overtones are integer multiples of the heard frequency, and there are no other voices being sounded by that particular speaker, then there will be no intermodulation distortion or frequency modulation distortion, but there will be a waveform distortion, altering the relative amplitude of the overtones, and thereby altering the sound as perceived by people.
Fortunately, the human ear does not care much about the phase of any given overtone. The various hair cells in the cochlea react selectively to various frequencies, but they do not keep track of the phase. This means that the phases of the overtones can be shifted to avoid instantaneous jumps in their sum. The CAD software should check for this and present workable waveforms to the speakers, while preserving the sound chosen by the user.
A Fourier transform is a function that represents a function of time (such as a waveform) by a sum of some number of sinusoidal functions. The phase is included, but the phase relationships need not be preserved in order to represent the sound as heard by people.
Yes, you post your proposal to this board on the forum: Adding Features - Audacity Forum
The proposal will remain on that board so that other Audacity users can comment, discus agree or disagree, then about 1 month after the discussion has concluded, one of the forum crew will transfer the feature request to the feature request page of the Audacity wiki.
Note that the proposal should describe the requested feature in a clear way that can be summarised on the wiki.