brudgers a day ago

If you want Midi to render consistently across devices, you are “Significantly Out of Luck.”

The audio sounds like it sounds because Microsoft Licensed Roland’s GS Wavetable. Without that you lose timbral information.

  • alexjplant 19 hours ago

    I remember being _very_ frustrated as a child that the music in "Hover!" sounded significantly different when we upgraded from our 486 to a Pentium II machine. The Sound Blaster gave the music a very distinct quality that was lacking in the software MIDI synthesizer.

  • Kwpolska a day ago

    That’s not the Microsoft/Roland sound.

    • brudgers a day ago

      Canyon.mid on my windows computer is.

      • Kwpolska a day ago

        It sounds noticeably different in Windows Media Player. The video is probably using a Yamaha OPL chip, as found in Sound Blaster cards.

        • brudgers a day ago

          The video is using audio because of variable Midi rendering.

          And any OPL is almost certainly emulated, not an actual chip.

rtkwe a day ago

That's probably mostly audio too given the vido is a 95% static screen. That has to encode to basically nothing.

  • seba_dos1 a day ago

    It's 2/3 audio and 1/3 video - 1.1MB.

    Perhaps somewhat surprisingly, most video codecs aren't tremendously efficient with static screens.

    • rtkwe 10 hours ago

      I'm not surprised. Generally the default number of frames between key/I-frames is set pretty low because too many I frames in a row tend to look bad on video content that isn't static and also if you jump to a point between 2 I frames the video can look really weird becasue it hasn't updated the I frame the P frames are manipulating.

    • jl6 a day ago

      The codec can be tuned to use a very long keyframe interval (e.g. ffmpeg’s -g option), but whether that’s worth the effort is another question.

agos 16 hours ago

a good quality soundfont is easily more than 3.4MB! But maybe back then nobody was using a soundfont that big