BPM: README for bpm

General

Get audio files as .mp3 from YouTube:

NOTE:

The youtube-dl tool has been blocked, but there is a fork: yt-dlp

yt-dlp -x --audio-format mp3 -o "feels.%(ext)s" https://www.youtube.com/watch?v=ozv4q2ov3Mk
yt-dlp -x --audio-format mp3 -o "pump-up-the-jam.%(ext)s" https://www.youtube.com/watch?v=9EcjWd-O4jI
yt-dlp -x --audio-format mp3 -o "blue-monday.%(ext)s" https://www.youtube.com/watch?v=c1GxjzHm5us
yt-dlp -x --audio-format mp3 -o "lady-bump.%(ext)s" https://www.youtube.com/watch?v=NelzIKT9ON4
yt-dlp -x --audio-format mp3 -o "girl-you-know-its-true.%(ext)s" https://www.youtube.com/watch?v=RdSmokR0Enk

Get the same files as .wav:

yt-dlp -x --audio-format wav -o "feels.%(ext)s" https://www.youtube.com/watch?v=ozv4q2ov3Mk
yt-dlp -x --audio-format wav -o "pump-up-the-jam.%(ext)s" https://www.youtube.com/watch?v=9EcjWd-O4jI
yt-dlp -x --audio-format wav -o "blue-monday.%(ext)s" https://www.youtube.com/watch?v=c1GxjzHm5us
yt-dlp -x --audio-format wav -o "lady-bump.%(ext)s" https://www.youtube.com/watch?v=NelzIKT9ON4
yt-dlp -x --audio-format wav -o "girl-you-know-its-true.%(ext)s" https://www.youtube.com/watch?v=RdSmokR0Enk

Get audio file with MSYS2:

pacman -S mingw-w64-x86_64-ffmpeg
wget https://github.com/yt-dlp/yt-dlp/releases/latest/download/yt-dlp.exe
./yt-dlp.exe --ffmpeg-location /mingw64/bin/ffmpeg.exe -x --audio-format wav -o "faulty-towers.%(ext)s" https://www.youtube.com/watch?v=PseBNmbIzGg

Play an audio file on Linux:

play pump-up-the-jam.mp3

Play an audio file on MS Windows - external soundcard (0), internal speakers (1), monitor speakers (2):

sox pump-up-the-jam.mp3 -t waveaudio 0
sox pump-up-the-jam.mp3 -t waveaudio 1
sox pump-up-the-jam.mp3 -t waveaudio 2

Play with MSYS2 - internal speakers (0), monitor speakers (1):

pacman -S mingw-w64-x86_64-sox
/mingw64/bin/sox.exe faulty-towers.wav -t waveaudio 0

Expected results:

feels.mp3: BPM according to other sources: 101, 101, 101
pump-up-the-jam.mp3: BPM according to other sources: 125, 128, 124, 125, 61
blue-monday.mp3: BPM according to other sources: 130, 126, 130, 130, 128
lady-bump.mp3: BPM according to other sources: 122, 122, 122, 122

Goals

Acceptable: Analyze an entire audio file, get a single BPM result (batch job).
Example: input: MP3 file, output: 125 BPM
Good: Read audio from a stream, detect BPM in real time.
Example: input: stdin, output: {125, 133, 125} BPM
Perfect: Read audio from a stream, detect BPM changes in real time, indicate in which bars (measures) changes occur, as well as in absolute time.
Example: input: stdin, output: {{bpm:125, bar:1, time:0.00}, {bpm:133, bar:10, time:6.00}, {bpm:125, bar:25, time:11.00}}

Conclusion

Use:

Use aubio (complete library with various tools, among them a beat detector) for real-time detection, reading from files, pipes, etc, using an FFT algorithm from 2005.
Use BTrack (small collection of C files) if you want to start from scratch. Good for real-time detection, using an FFT algorithm from 2011.
Use beat-detector if you know Rust.
Use bpm from the Linux package bpm-tools if nothing else is available, or if you are lazy. Applying a multiplying factor may improve results.

Results

`read-from-aubiotrack`

This program uses the output from aubiotrack and calculates both the median and the average values.

    make clean all

    aubiotrack -i feels.wav | ./bin/read-from-aubiotrack
    DEBUG: stored numbers = 381
    DEBUG: read errors = 0
    DEBUG: BPM median value: 102.222836
    DEBUG: BPM average value: 106.508915

    aubiotrack -i pump-up-the-jam.wav | ./bin/read-from-aubiotrack
    DEBUG: stored numbers = 451
    DEBUG: read errors = 0
    DEBUG: BPM median value: 126.532626
    DEBUG: BPM average value: 125.800103

    aubiotrack -i blue-monday.wav | ./bin/read-from-aubiotrack
    DEBUG: stored numbers = 977
    DEBUG: read errors = 0
    DEBUG: BPM median value: 131.987065
    DEBUG: BPM average value: 131.713718

    aubiotrack -i lady-bump.wav | ./bin/read-from-aubiotrack
    DEBUG: stored numbers = 359
    DEBUG: read errors = 0
    DEBUG: BPM median value: 123.599973
    DEBUG: BPM average value: 122.601263

    aubiotrack -i girl-you-know-its-true.wav | ./bin/read-from-aubiotrack
    DEBUG: stored numbers = 394
    DEBUG: read errors = 0
    DEBUG: BPM median value: 98.764457
    DEBUG: BPM average value: 102.734908

Conclusion:

aubio tempo uses the average of all values generated by aubiotrack to calculate the BPM.
Anyhow, the median value is closer to the expected value in most cases.

README for bpm

General

Goals

Conclusion

Results

read-from-aubiotrack

Details/README about each external project

`read-from-aubiotrack`