It's a seemingly simple question with a somewhat complex explanation.

MIDI does not specify the actual sound to be played, only the information on what to do with the sound, like note, pitch, vibrato, etc, and which instrument to use, like piano. It is up to the playback device to decide what the sound to be played actually sounds like.

General MIDI is the most basic set of information a MIDI device must support to be classified as a MIDI device. The MIDI standards were developed a "long time ago" (early 1980's).

Sometime afterwards (later 80's) Yamaha and Roland wanted additional features for their devices, and as such, each developed their own extensions to the MIDI standard. Yamaha made XG, and Roland made GS. Both provide many more available instruments and variations of instruments than the original General MIDI.

Devices that support more than one standard, like your Yamaha keyboard, usually have a "standard sound set" (called a bank). It is usually a basic set. Then better/alternate sound sets can be accessed/used by selecting the sound sets from in the device (or from in a MIDI file).

GM is limited to 128 instrument sounds (more if using the General MIDI 1 Level 2 standard [GM2, not to be confused with the new MIDI 2.0 specification currently in development]) and... ummm, I forget how many percussion sounds. All GM devices should support these sounds, accessed the same way between all devices.

The XG and GS extensions add tons of extra sounds and variations, but are only supported by devices specifically designed to support them, and they use different methods of bank selection to access these extra sounds and sound modifications.

You can hear some of the differences by using a MIDI player on a computer, where you can select from a massive array of sound banks to play back your files.

I personally have quite a few different sound sets (in the sF2 soundfont format) on my PC to play back my files either as realistic sounding instruments or as a variety of older sound chips, including Nintendo NES and SNES, Sega Genesis, OPL3 (old Soundblaster PC sound card), and even specifically the Yamaha SYXG50. Each one of these has a completely different sound from the others, and each one is General MIDI compliant. The actual MIDI files are tiny and only tell it what to do, not what the sounds are. It is up to my MIDI player to provide these sounds.

I have a basic understanding of MIDI. And I've read its protocol specification. I'm also a software developer, and I want to explore on MIDI as a hobby.

But my question is: Why can't manufacturers have the 128 GM sounds as base foundation, with a quality of built-in sounds and all the rich enhancements they can made with the sound via its powerful oscillators, effects, filters, modulation etc. That means, only one piano which is MIDI GM compliant and still sounds natively rich with full power of the synthesis processing power of the keyboard has. In cases where a manufacturer has additional sounds apart from what GM MIDI standardizes, then there is no problem it can just expand at the end of what the GM MIDI standard mandates. That means, add proprietary sounds starting at 129 and up.

But what is happening instead is the following: There is a GM MIDI piano sound. And then, there is a native proprietary rich piano sound. Why can't manufacturers blend or combined and have 1 piano per the GM MIDI standard, and yet, can still be a very rich piano sound with full synthesis enhancements. Why have two pianos, wherein the GM MIDI piano version is being excluded from a richer synthesis. And then, there is a native piano which is the piano with richer synthesis?

Is there a specific aspect of the MIDI protocol that is lacking and that is causing this divide?

Please note that this is nothing to do with MIDI. The original question pointed the finger at the Manufacturers, which is correct, so please keep that in mind.

The Manufacturers clearly like to include/add as many fancy bells and whistles to their devices as they can, and make then as sophisticated as they can (all within an appropriate budget, I assume) but they also wish to allow for the less sophisticated users so they include a GM option and they try to keep that as simple as possible.

A super user can of course easily change a midi file to replace the basic GM sound with one of the more exotic sounds from the device, but then playback of the midi file will no longer be compatible on different GM devices. Maybe slightly, maybe by a lot. This partly undermines the whole idea of GM.

Maybe at some time GM could be implemented as a table which the user MAY edit. The table would be provided with a default setting to a standard GM sound set, so that the unit remains ready to go out of the box. But the user could then amend the table, and maybe the unit could remember the new settings for future use. The changes could swap the standard sound for any better ones available.

This might help the complaint being made here, which I DO understand. This will however be totally up to the individual manufacturers.

Geoff

The idea behind the GM specification is compatibility.

If you author a GM SMF it should sound broadly the same no matter whose hardware/software you play it on.
The same with GM2.
Any effects used e.g. chorus or reverb, will work in the same fashion and the file will sound the way it should no matter which manufacturer.

JohnG.

One other thing to note is that because the standard is based on compatibility, and is more than three decades old, many devices simply wouldn't be able to store or play the "fancier" sounds due to hardware limitations. Many devices had very limited storage space, and one of the benefits of MIDI is the very small byte size of the standard and the files/data that relates to it. Many of the "fancier" sounds are based on sound samples, which are orders of magnitude larger in size than a few bytes of code. Imagine a text file that lists the notes played in a song, and then compare it to an mp3 of that song. The absolute largest MIDI file I have is less than 500kB, and that is an entire game's soundtrack more than an hour long, in the GS format (so much larger than a GM file would be). The playback device would only receive a few bytes of that at a time. In order for the file to sound exactly the same on all devices, either the song or all devices would need a set of sound samples that covered the complete range of General MIDI (plus all the extras that GS supports). The MIDI standard does not support transfer of sound samples, only data, therefore it is up to the device to provide the samples. If the sound chip or the storage on the device simply can't hold that large set of sound samples, the only option is synthetic sound (hence why they are known as synthesizers). And then you can only produce the types of sounds the chip/device is capable of making. So Roland (in this case) wants to use the fancier sounds they have available, which requires selecting sample banks that fall outside of the standard GM sound set. Many of the features to give that richer sound do not fall in the GM standard, and are part of the GS or XG extensions, so the GM sounds they provide are not as rich and nice sounding without turning on all the extra bells and whistles.

That being said, if the past were to suddenly disappear, and MIDI was made today with no need for compatibility with all that is already out there, I feel that having the richer sounds by default is EXACTLY what they would do to entice you to buy their particular devices 🙂