General MIDI Working Group (GMWG)

Report on Clarifications to GM Level 1

The following items were parts of the opening mission statements of the GMWG:

• "The goal of this working group is to make recommendations to the Interactive Audio SIG and the MMA to help encourage the manufactures of General MIDI instruments that will play back General MIDI files in a musically satisfactory way regardless of the instrument for which the file was composed."

  "We pursue greater compatibility for several reasons: consumers will know that their hardware and software will work together, composers and software developers will be able to support all instruments using only one GM file. Instrument manufacturers will be able to boast true compatibility with GM music. The industry will see more sales because of consumer and developer confidence in the viability of GM."

  "There is considerable discussion among manufacturers about the intentional gaps in the GM spec (which is currently a set of recommended practices.) On the up side, these gaps give freedom to individual manufacturers to create whatever instruments and timbres they choose. The down side is that GM compatibility is questionable."

The GMWG was not able to find a complete solution for reliable GM1 compatibility. We had hoped to make some recommendations for how to implement GM1, or provide some higher level definitions for all aspects of GM1 that can cause incompatibilities.

We prepared a list, "Specific Issues of GM Compatibility to Clarify" that outlined areas where GM1 devices can be incompatible. For most of these items we thought to have been able to provide guidelines based on clear technical recommendations. However, our discussions lead us back to the our biggest incompatibility concern, which we ended up calling "The Great Dilemma". It seemed that unless we solved this central concern any solutions we suggested for other issues would not provide substantial improvements in compatibility.

George Sanger wrote:

• The Great Dilemma
  • The biggest problems of incompatibility between GM instruments revolve around Tonality and relative Volume of each tone. I see two possible methods to achieve greater compatibility in these issues between GM instrument systems:
    • Method 1: -Instrument sounds must be restricted to being very similar in tonality to a reference set. The volume/ADSR is specified scientifically by dB.
    • Method 2: -Instruments may have any tone. The volume/ADSR must be perceived to be very similar to a reference set.

In either method above a Reference Model and/or Technical Definition is needed for the Waveform Sound Set and/or Amplitude Envelope and Relative Tone Volume. Unfortunately, it is extremely difficult if not impossible to define a standard set for these issues by any mathematical formula or technical description. For example: If amplitude and envelope are defined, the perceived volume and envelope are still affected by the changing harmonic content of an undefined waveform.

Also, the group considered that there are many GM1 products, hardware and software, already in the marketplace. Our group did not think it would be feasible to define any recommendations that would render the large body of existing products as incompatible.

Therefore, without a technical definition to solve "The Great Dilemma" of GM1 compatibility concerns, and without a reference model for waveforms and envelopes, the General MIDI Working Group could not provide any absolute technical recommendations to enhance compatibility between various GM1 instruments.

However, the group's discussions did help identify which areas of compatibility are most important to game developers, and this information will be very useful in later discussions of GM extensions.

It was also clear that developers and hardware manufacturers are intent upon maximizing compatibility, regardless of the existing technical obstacles, and for that reason the vast majority are using a Roland Sound Canvas as their reference when creating GM products.

Building on that fact, the working group did recommended that all GM devices use a MIDI volume and MIDI expression response curve which matches that of the Sound Canvas, and was reported by Yamaha to also be used in their GM products (see below). Still, this conclusion is more an acknowledgment of current conditions than it is any statement of artistic preference ... the latter being the preferred outcome of this group.

For more details of the discussions of the GMWG, an archive of all email messages is available from the IASIG. This report ends the work of the General MIDI Working Group. A new group will be meeting to discuss the specific issues of General MIDI 2 which would interest interactive developers.

Considerations for the Future use of MIDI in Interactive Audio Applications

The GMWG efforts of the past year were focused primarily on GM1. However, the group also brought up some other related discussions that are worth considering. Following are some of the other interests of the members of the GMWG. We hope that these ideas will be useful for other working groups of the IASIG such as the Downloadable Samples group and discussion in a GM Level 2 working group.

Downloadable Samples

• It seems that among highest priorities of most GMWG members for the future application of MIDI is the need for an industry wide support for a Downloadable Samples Standard. This may provide better reliability for compatibility than currently available in General MIDI. This also will provide the opportunity for content developers to differentiate their products in the marketplace. We look forward to a standard being set by the MMA.

General MIDI Level 2

• The GMWG believes that for any future definition of General MIDI such as GM2 to be most successful it should be a more highly defined standard than GM1. This will help to avoid some of the incompatibility concerns that continue with GM1 products. GM2 products may face what the GMWG refers to a the The Great Dilemma that we found when we tried to solve GM1 compatibility concerns.
• GM2 development should be closely related to any future definition of a Downloadable Samples Standard. Possibly GM2 could incorporate a support mechanism for the DS Standard. Or a DS Standard could possibly be an integral part of GM2. If the DS Standard is separate from GM2 it seems that GM2 devices that include a DS feature should support the DS Standard.
• The GMWG would like to have more musical control over the sounds in a GM2 device. They would like to be able to have easy access to alter or edit parameters of the sound such as envelope attack, decay, sustain, and release times, filter cutoff, vibrato speed, etc. This access should be provided using an easily implemented system such as controllers like NRPN as now found on a number of GM1 devices or by newly defined RPN.
• The GMWG would like GM2 be as highly compatible to GM1 and existing GM1 products (hardware, software, and Content) as is technically feasible.
• The GMWG has expressed interest in having Reverb and Chorus included as requirements in the specification of GM2.
• For MIDI implementation outside of the requirements of GM2, we would like GM2 products to use MIDI in the most natural way as defined within other parts of the MIDI specifications. For example, even if GM2 did not define a requirement for Sostenuto on controller #66, GM2 devices should not freely use controller #66 for any function other than Sostenuto.
• The GMWG has expressed interest in having the GM2 Specification include guidelines for GM2 content. GM1 was only a GM device definition. A GM2 definition that includes both device design and content design may help all parties to work toward the best compatibility.

The GMWG has expressed concerns about the sound set of GM1. For example, there were differing opinions whether the French horn in most GM devices was suitable. The reasonable solution seemed to be to have various French horn sounds for different applications. The GMWG would like a larger sound set than in GM1, expanding somewhat on the basic 128 sound set by providing variation sounds.

-Submitted by George Sanger & Mike Kent, Co-chairmen