Configuration#

DrumScript’s DSP behaviour is controlled by module-level constants rather than per-call parameters. This page documents what is actually configurable today.

Note: Earlier versions of this page documented a threshold argument on detect_onsets() and a format argument for output. Neither exists. The real controls are described below.

Onset Detection#

detect_onsets() takes the audio and its sample rate, and nothing else:

import drumscript as ds

y, sr = ds.load_audio("drum_audio.wav")
y = ds.normalise_audio(y)
onsets = ds.detect_onsets(y, sr)   # returns a list of times in seconds

Sensitivity is governed by two values set inside drumscript/audio_processor/onset_detector.py:

Setting	Value	Effect
`delta`	`0.05`	Detection threshold above the noise floor. Lower detects quieter hits (more false positives); higher misses ghost notes.
lockout	`0.05` s	Minimum gap between two detected hits. Stops cymbal vibration from double-triggering.

Both are currently hardcoded. Exposing them as arguments is on the roadmap — see the issue tracker.

Sample Rate and Hop Length#

Set in drumscript/notation_generator/constants.py:

Constant	Value	Meaning
`SAMPLE_RATE`	`44100`	Rate the pipeline resamples to internally.
`HOP_LENGTH`	`128`	Analysis window advance, ~2.9 ms at 44.1 kHz.

load_audio() defaults to the file’s native rate. The pipeline passes sr=SAMPLE_RATE explicitly when it needs a consistent rate:

from drumscript.notation_generator.constants import SAMPLE_RATE

y, sr = ds.load_audio("drum_audio.wav")                 # native rate
y, sr = ds.load_audio("drum_audio.wav", sr=SAMPLE_RATE) # resampled to 44100

Output Formats#

Output format is not a parameter. transcribe() always writes three files and reports where they went:

result = ds.transcribe("drum_audio.wav")
result["pdf_path"]    # sheet music
result["json_path"]   # raw transcription data
result["midi_path"]   # MIDI for DAW import

MusicXML is not produced automatically. Export it separately with ds.export_xml().

Per-format flags (output_midi, output_json, output_xml) are planned — see the CHANGELOG.

Stem Output Format#

The one place a format is selectable is stem export, via --format on the CLI or output_format= in the Python API:

drumscript "full_song.mp3" --all-stems --format mp3

ds.extract_stems("full_song.mp3", all_stems=True, output_format="mp3")

Valid values are wav (default) and mp3. MP3 requires ffmpeg; WAV does not.

Time Signature#

ds.transcribe("drum_audio.wav", time_signature="6/8")

drumscript "drum_audio.wav" --ts 6/8

Use a forward slash (3/4, 6/8). Any other form — including underscores like 3_4 — silently falls back to 4/4 with no warning.