MIDI itself, which stands for "Musical Instrument Digital Interface", is just a protocol and standard for communication between musical instruments and any device that understands the language. It can be used to control an array of synthesizers, make a tin can sound like a drum, or even operate industrial equipments.
The scope of this article, however, will mainly focus on the usage of MIDI in computer systems for playback of files that contain MIDI data. These files usually come with the .mid extension, and were hugely popular in the golden days of multimedia computing to share music. In professional music composition/arrangement, it still plays a vital role.
Without going into the details of what the format is composed of, you just need to understand that a MIDI file eg. foobar.mid does not contain any digital audio data, hence no "PCM stream". It is a common misconception that MIDI is a sound file format, and as such you usually see people complaining that music players like Amarok cannot play the file. Here is a very newbie-friendly outline of what a MIDI/MID file contains:
# FOOBAR.MID Note ON Use Instrument #1 Play Note C1 Set Volume at 100 Set Pitch at 50
In order for such a file to be useful, there needs to be an "engine" that can translate the data to music. This engine will have a "tone generator", and this is what we call a "synthesizer". So any player that can play back a MIDI file without MIDI-capable hardware (your computer's sound device), has a synthesizer built-in or uses an external one. A typical keyboard (not the QWERTY thing you are typing on) is actually made up of two components - a MIDI "controller" (the keys) and a synthesizer (tone generator/module; the thing that makes sound).
So up to this point, you should be able to understand that:
- There needs to be a synthesizer to play a MIDI file.
- A synthesizer can be hardware or software.
- Most computer soundcards/chipsets do NOT have synthesizers.
- You need a synthesizer with a proper "bank" (collection of sounds) to be able to enjoy all the glory of MIDI files.
- If a certain instrument is not in the bank, your synthesizer will not play anything for notes using that instrument.
- If a certain instrument in the file corresponds to a different one in the bank, your synthesizer will play a different sound (obviously).
General MIDI (GM) is a specification to standardise numerous MIDI-related matters, particularly that of instruments layout in a collection of sounds. A "soundbank" which is GM-compatible means that it meets the criteria of General MIDI, and as long as the MIDI file is also GM-compatible (as in nothing extraordinary is defined - such as introducing a new instrument or having one in a different location of the bank), the playback will be as intended since the bank has the correct instrument/handler for the MIDI message/event. One of the most popular soundbank formats is that of SoundFont, particularly SF2.
- If you have a soundcard which can make use of soundfonts, you can load a .sf2 file onto it.
- If you do not have a soundcard which can make use of soundfonts (basically no hardware synthesizer), you can use a software synthesizer and load the SF2 file. In turn, you can find some way to globally make use of this synthesizer.
"Why can I play MIDI with Windows Media Player, then?"
Well, because Windows has a default software synthesizer which acts globally. Even then, it lacks the quality which should be expected of modern computers. If there were a way to do it on Linux, you would be able to play back MIDI from any player too. Perhaps a MIDI server (which will hold a synthesizer of choice like TiMidity++ or FluidSynth) that sits within the sound server, like Phonon or PulseAudio. Nevertheless, nothing of this sort has been implemented and you can only play MIDI with a player that has a plug-in to source a synthesizer or has a synthesizer itself. Such an example is XMMS.
(Details on soundcards and MIDI, possibly link to SBLive MIDI here..)
(Details on options available, like TiMidity++ or FluidSynth. Can be merged or linked if page exists, eg. there's an article for Timidity already.)