How to convert MIDI file into audio file and preserve tempo?

This is a bit of a musical question.

First, I generate a click track in MIDI at 1/8 = .1250 and save the data in a MAT file. I don't need the whole MIDI header etc. since I just need to use the data to create an audio file. Since I don't specify a tempo in BPM, I assume it is 120.

Next I attempt to convert the MIDI info into an audio file, using Sine waves as the clicks. When I do this, the track is not at the same tempo as the MIDI file. Why is this?

(Excuse my clumsy programming.)

Fs = 96000; % SAMPLE RATE OF AUDIO FILE

% CREATE SINE WAVE PULSE
sine1 = dsp.SineWave();
sine1.SamplesPerFrame = Fs;
sine1.Frequency = 5;
fSine = sine1();

% CREATE AN EMPTY 1-SEC WAVE FILE TO HOLD OUTPUT
zOUT = zeros(Fs*1,1); % NUMBER OF SECONDS

% LOAD MIDI FILE INFO AND PUT IT INTO SEPARATE ARRAYS
zF = load("Click Track TEST.mat"); % LOAD MIDI FILE INFO
cMALL = zF.cZ;
cMTRACK = cMALL(:,1);
cMPITCH = cMALL(:,3);
cMVEL = cMALL(:,4);
cMDb = cMALL(:,5); % BEGIN DURATION FOR EACH PULSE
cMDe = cMALL(:,6); % END DURATION
cMsiz = size(cMTRACK,1);

aa = 0;
zz = 0;
cc = 1;
for ii = 1:cMsiz
    % MIDI DURATION CALCULATE
    bD = cMDb(ii); % MIDI DURATION BEGIN
    eD = cMDe(ii); % MIDI DURATION END
    cMDz = eD - bD; % DURATION OF THE PULSE
    % CONVERT MIDI DURATION TO AUDIO FRAME # DURATION
    cMDFsz = round(cMDz * Fs); % CONVERT MIDI TO SAMPLE DURATION
    zzend = cMDFsz;
    % ALTERNATE BETWEEN SINE WAVE AND SILENCE TO HEAR PULSE
    if mod(ii,2) == 1
        cMP = fSine(1:cMDFsz);
    else
        cMP = 0;
    end
    zOUT(zz+1:zz+zzend,1) = cMP;
    zz = zz + zzend;
end
 
% WRITE FILE AS WAVE AUDIO
fileOUT = ' Click Test 96K 2023.wav';
audiowrite(fileOUT,zOUT, Fs);

Answer 1

[Binaya]

Hi Mark

I understand that you would like to know why the audio file generated from MIDI file is not playing at the same tempo as the original MIDI track.

The main reason of the tempo mismatch can be:

1. BPM: If not specified, the assumed BPM of 120 may differ from the track's actual tempo. With MIDI clicks at 0.1250s and assuming there are two notes played for each click, the resulting tempo is 240 BPM.
2. Sampling rate: The sampling rate used in the code can be different from the original sampling rate.

I hope this helps.

Regards

Binaya

Answer 2

[Voss]

zF = load('Click Track TEST.mat'); % LOAD MIDI FILE INFO
cMALL = zF.cZ;
disp(cMALL)

    1.0000    1.0000   75.0000   96.0000         0    0.1250
    1.0000    1.0000   60.0000   96.0000    0.1250    0.2500
    1.0000    1.0000   60.0000   96.0000    0.2500    0.3750
    1.0000    1.0000   60.0000   96.0000    0.3750    0.5000
    1.0000    1.0000   75.0000   96.0000    0.5000    0.6250
    1.0000    1.0000   60.0000   96.0000    0.6250    0.7500
    1.0000    1.0000   60.0000   96.0000    0.7500    0.8750
    1.0000    1.0000   60.0000   96.0000    0.8750    1.0000
    1.0000    1.0000   75.0000   96.0000    1.0000    1.1250
    1.0000    1.0000   60.0000   96.0000    1.1250    1.2500
    1.0000    1.0000   60.0000   96.0000    1.2500    1.3750
    1.0000    1.0000   60.0000   96.0000    1.3750    1.5000
    1.0000    1.0000   75.0000   96.0000    1.5000    1.6250
    1.0000    1.0000   60.0000   96.0000    1.6250    1.7500
    1.0000    1.0000   60.0000   96.0000    1.7500    1.8750
    1.0000    1.0000   60.0000   96.0000    1.8750    2.0000
    1.0000    1.0000   75.0000   96.0000    2.0000    2.1250
    1.0000    1.0000   60.0000   96.0000    2.1250    2.2500
    1.0000    1.0000   60.0000   96.0000    2.2500    2.3750
    1.0000    1.0000   60.0000   96.0000    2.3750    2.5000
    1.0000    1.0000   75.0000   96.0000    2.5000    2.6250
    1.0000    1.0000   60.0000   96.0000    2.6250    2.7500
    1.0000    1.0000   60.0000   96.0000    2.7500    2.8750
    1.0000    1.0000   60.0000   96.0000    2.8750    3.0000
    1.0000    1.0000   75.0000   96.0000    3.0000    3.1250
    1.0000    1.0000   60.0000   96.0000    3.1250    3.2500
    1.0000    1.0000   60.0000   96.0000    3.2500    3.3750
    1.0000    1.0000   60.0000   96.0000    3.3750    3.5000
    1.0000    1.0000   75.0000   96.0000    3.5000    3.6250
    1.0000    1.0000   60.0000   96.0000    3.6250    3.7500
    1.0000    1.0000   60.0000   96.0000    3.7500    3.8750
    1.0000    1.0000   60.0000   96.0000    3.8750    4.0000

Assuming 120 BPM (i.e., 1 beat every 0.5 seconds), then each 4 rows is one beat.

A guess: You shouldn't be turning on the sine wave every alternate row:

if mod(ii,2) == 1

cMP = fSine(1:cMDFsz);

else

cMP = 0;

end

but every fourth row, based on where the 75's are in the third column

if cMPITCH(ii) == 75

cMP = fSine(1:cMDFsz);

else

cMP = 0;

end

Here's your code, incorporating that change (and re-written in a more concise style), including a plot that shows you get one duration-0.125-second click each 0.5 seconds (= 8 clicks in 4 seconds = 120 BPM).

Fs = 96000; % SAMPLE RATE OF AUDIO FILE

% CREATE SINE WAVE PULSE
sine1 = dsp.SineWave();
sine1.SamplesPerFrame = Fs;
sine1.Frequency = 5;
fSine = sine1();

zF = load('Click Track TEST.mat');
zOUT = zeros(Fs*zF.cZ(end,6),1);
zz = 0;
for ii = 1:size(zF.cZ,1)
    N = round((zF.cZ(ii,6) - zF.cZ(ii,5)) * Fs);
    if zF.cZ(ii,3) == 75
        zOUT(zz+1:zz+N,1) = fSine(1:N);
    end
    zz = zz + N;
end
 
% WRITE FILE AS WAVE AUDIO
fileOUT = ' Click Test 96K 2023.wav';
audiowrite(fileOUT,zOUT, Fs);

plot((1:zz)/Fs,zOUT)