phasefinder is a beat estimation model that predicts metric position as rotational phase, heavily inspired by this paper.

full writeup here

usage

python

from phasefinder import Phasefinder

# initialize model
pf = Phasefinder()

# predict beats
audio_path = "path/to/your/audio/file.[wav/mp3/flac/etc]"
beat_times = pf.predict(audio_path)

# predict beats and BPM
beat_times, bpm = pf.predict(audio_path, include_bpm=True)

# generate a click track
output_path = "output_with_clicks.wav"
pf.make_click_track(audio_path, output_path, beats=beat_times)

cli

basic usage

python -m phasefinder.infer path/to/your/audio/file.wav

This will print the estimated beat times to the console.

options

• --bpm: Include BPM in the output
• --noclean: Don't apply the cleaning function to the beat times
• --format {times,click_track}: Choose the output format (default: times)
  • times: Output beat times
  • click_track: Generate an audio file with click track
• --audio_output PATH: Specify the path for the output audio file with clicks (default: output_with_clicks.wav)
• --json_output PATH: Save the results to a JSON file

examples

1. Estimate beats and BPM:

   python -m phasefinder.infer path/to/audio.wav --bpm

2. Generate a click track:

   python -m phasefinder.infer path/to/audio.wav --format click_track --audio_output output.wav

3. Save results to a JSON file:

   python -m phasefinder.infer path/to/audio.wav --bpm --json_output results.json

4. Estimate beats without applying the cleaning function:

   python -m phasefinder.infer path/to/audio.wav --noclean