Holy cow that MT-32 MIDI is crazy. The sound design is incredible! Thank you for sharing! It's nice to hear a snare on the MT-32 that isn't the stock PCM one. 😁

It's a crazy midi. I found the original here:
https://www.youtube.com/watch?v=cTraSI6LMaI

It's an RCP and CM6 (it's in the video description), which you can convert to MIDI with this tool:
https://shingo45endo.github.io/rcm2smf/

Interestingly enough, the converter has a setting relevant to the SysEx discussion:

Interestingly, I noticed that in mt32emu the scrolling text on the display keeps getting interrupted by the MIDI activity flashing the display back to the channels view and back and forth every time the text scroll is updated. But on a hardware MT-32 this doesn't happen.

That does happen on mine (text display constantly interrupted), which is MT-32 old. Your's is probably a "new"?

MusicallyInspired wrote on 2026-06-04, 14:28:

Interestingly, I noticed that in mt32emu the scrolling text on the display keeps getting interrupted by the MIDI activity flashing the display back to the channels view and back and forth every time the text scroll is updated. But on a hardware MT-32 this doesn't happen.

What do you mean by "mt32emu" exactly? If you mean the mt32emu-qt.exe GUI app then the answer is the display behavior is ROM dependent.
At the very beginning of Synth.cpp you can find the ControlROMFeatureSet definitions. The relevant fields are "quirkDisplayCustomMessagePriority" and "oldMT32DisplayFeatures".
https://github.com/munt/munt/blob/master/mt32 … u/src/Synth.cpp

Okay. Why is it then that no matter which Control ROM I try it still happens? I do see that in MuntVSTi it behaves properly.

MusicallyInspired wrote on 2026-06-05, 15:58:

Okay. Why is it then that no matter which Control ROM I try it still happens? I do see that in MuntVSTi it behaves properly.

1. I do not know but it should not. Last time I tried it worked according to the configured settings.
You should make sure that on the synth's property page the relevant setting is set properly:

2. It's because in MuntVSTi I always treats both relevant fields as false. Otherwise there would be glitches in the text rendering.
There is an API call provided by the Munt library that enables such overrides:

1**
2 * Permits to select an arbitrary display emulation model that does not necessarily match the actual behaviour implemented
3 * in the control ROM version being used.
4 * Invoking this method with the argument set to true forces emulation of the old-gen MT-32 display features.
5 * Otherwise, emulation of the new-gen devices is enforced (these include CM-32L and LAPC-I as if these were connected to an LCD).
6 */
7MT32EMU_EXPORT_V(2.6) void MT32EMU_C_CALL mt32emu_set_display_compatibility(mt32emu_const_context context, mt32emu_boolean old_mt32_compatibility_enabled);

BTW, I have not met any cases where the 'true' state of those flags resulted in any benefits.

polynektar wrote on 2026-06-03, 17:24:

One more note: Free Rein causes a "buffer overflow" on my MT-32 Old. I extracted the SysEx, and sent it slower, thru Falcosoft Midi Player, and it didn't cause the overflow, then played the song. Not sure if anything can be done to prevent the overflow, otherwise.

I've tested the file today with different players and connection methods and haven't yet been able to replicate the buffer overflow with my MT-32 old. The 7kb or so of SysEx data is all encoded at 0:00:00, so the lack of any observable issues on my end might be wholly down to my MIDI interfaces (a Roland UM-ONE USB interface on my Win11 PC and an ESS Solo-1 with gameport/MIDI adapter on my Win98 PC) doing a good job of buffering and preventing data overflow.

I did, however, find that running the MIDI file through Windows 98 Media Player causes the first few notes to be missed, as it's presumably starting to send note data (which also starts at 0:00:00) before the SysEx data has finished being transmitted.

I've made another version of the file, shifting all the note data to begin playing from the second bar. I've also added approx. 50ms delay after each SysEx block by using a hex editor to run a global Find & Replace from "F7 00" to "F7 60". Hopefully this should improve compatibility with an MT-32 old on any possible combination of system or player.