Determine the total length of a midi file

[tqwewe]

I've created a function to convert the delta to miliseconds:

const microsecondsPerQuarter = (
  midi.tracks[0].find(({ setTempo }) => setTempo).setTempo
).microsecondsPerQuarter;

const deltaToMs = (delta, bpm = 110) =>
  (60_000_000 / (bpm * microsecondsPerQuarter)) * delta;

I'm not sure if this is even correct, but it seems to work for now.
#305 Says something about divisions but I'm not sure how to use that.

From here, how can I determine the total miliseconds of the midi track in order for a perfect loop?

[chrisguttandin]

Assuming you only have one setTempo event for the entire track you could do it like this.

seconds = delta * microsecondsPerQuarter / division / 1000000

[tqwewe]

Hi thanks for your reply!

Where does delta come from exactly?

My data looks like this:

{
  "division": 384,
  "format": 1,
  "tracks": [
    [
      {
        "timeSignature": {
          "denominator": 4,
          "metronome": 24,
          "numerator": 4,
          "thirtyseconds": 8
        },
        "delta": 0
      },
      {
        "setTempo": {
          "microsecondsPerQuarter": 545454
        },
        "delta": 0
      },
      {
        "endOfTrack": true,
        "delta": 0
      }
    ],
    [
      {
        "trackName": "Electric Piano",
        "delta": 0
      },
      {
        "programChange": {
          "programNumber": 0
        },
        "channel": 0,
        "delta": 0
      },
      {
        "noteOn": {
          "noteNumber": 59,
          "velocity": 50
        },
        "channel": 0,
        "delta": 0
      },
      {
        "noteOff": {
          "noteNumber": 59,
          "velocity": 0
        },
        "channel": 0,
        "delta": 96
      },
      ...
      {
        "endOfTrack": true,
        "delta": 0
      }
    ]
  ]
}

[chrisguttandin]

It's the delta of the event. Let's say you want to figure out the time of the noteOff event in your example. Then it would be:

seconds = 96 * 545454 / 384 / 1000000

In other words the first node stops after about 0.136 seconds.

[tqwewe]

Ah yep I managed to calculate that. But I'm curious to get the entire duration for a perfect loop. For example, if the last note is not at the end of the bar, then there would be some space after of silence

[chrisguttandin]

I guess there is no way to derive that information from the file. If you know that your file contains the events for 16 beats you could just use microsecondsPerQuarter * 16 as the length of the loop. But if the last note ends at beat 15 then this will be the end of the MIDI file.

[tqwewe]

If I can know how many beats my file contains, then that'd be perfect. But I cannot find how to get that information.

Importing the file into Ableton live, it detects the full length correctly.
I used this website to write the midi: https://onlinesequencer.net/
And Ableton was able to get the correct lenght of my midi track.

I think the information must be in the file somewhere.

[chrisguttandin]

What's the time difference between your last event in the MIDI file and the expected duration that Ableton detects?

[tqwewe]

I created a midi using onlinesequencer.net.

I marked with a red line where I'd expect the loop to occur, and this is also where Ableton puts the loop (at the end of bar 2):

[chrisguttandin]

Maybe Ableton just rounds to the next full bar. You can use the denominator of the timeSignature event to determine how many beats a bar has.

[tqwewe]

Hmm I'm still trying to figure out how to use the denominator to calculate this.

[chrisguttandin]

I'm not pretending that I understood the ins and outs of the timeSignature event. I guess it takes years to understand it completely. :-)

Anyway, as far as I understand a denominator of 4 says a bar has 4 beats. And the microsecondsPerQuarter says a beat is 545454 ticks. When we convert this to seconds using the formula from above we end up with about 136ms.

Given that a bar has 4 beats a bar is about 544ms long. So if we want to end up with a full bar we need to make sure the length is a multiple of 544ms or 2181816 ticks (= 545454 * 4 ticks).

By looking at the image from above I guess the accumulated delta of your last noteOff event is about 3409088 ticks (something close to 545454 * 6.25 ticks).

If you now pick the next multiple of 4 you end up with 8.

Does that make sense?

[tqwewe]

Thanks so much for this reply! I'm trying it out now. I'm just a little confused why you chose to include the noteOff 96 ticks in the equasion, I guess because it's the last note off in the track?

I'll let you know how I go