Using Vamp Plugins for Feature Extraction in Audacity

This article explores Vamp plugins and their application for acoustic feature extraction using the Audacity digital audio workstation. Unlike traditional audio plugins that modify or alter how a recording sounds, Vamp plugins analyze audio signals to generate structured, meaningful data such as pitch, rhythm, and timbre. By integrating these plugins into Audacity, users can transform raw audio tracks into visual representations and quantifiable metrics suitable for musicology, speech research, and audio analysis.

What Are Vamp Plugins?

Vamp is an open-source audio processing plugin system developed by the Centre for Digital Music at Queen Mary University of London. While traditional audio plugins (such as VST, AU, or LADSPA) operate on a "sound in, sound out" model to add effects like reverb, compression, or equalization, Vamp functions on a "sound in, data out" basis.

Vamp plugins take raw audio as input, calculate mathematical transformations over time, and output descriptive information. This information can take several forms:

Time-stamped events: Specific points in time denoting discrete occurrences, such as drum hits, note onsets, or chord changes.
Labelled intervals: Continuous sections of time classified by specific properties, such as a designated musical key, structural segments (verse/chorus), or detected speaker activity.
Feature curves: Continuous, numerical measurements plotted over time, such as fundamental frequency (pitch), energy levels, or spectral centroid values.

What Is Acoustic Feature Extraction?

Acoustic feature extraction is the process of converting complex, continuous sound waves into a set of discrete, measurable parameters. In computational audio analysis and Music Information Retrieval (MIR), raw audio contains too much unstructured data to interpret directly.

Feature extraction breaks down this data into low-level and high-level descriptors:

Low-level features: Spectral characteristics (e.g., spectral flux, brightness), envelope analysis, and energy distribution.
Mid-level features: Pitch contours, chromagrams (distribution of energy across 12 semitone pitch classes), and harmonic analysis.
High-level features: Semantic interpretations such as tempo estimation, beat tracking, melody extraction, and genre classification.

How Vamp Performs Feature Extraction in Audacity

Audacity includes native support for Vamp analysis plugins, leveraging them through its built-in analytical tools to display visual metadata alongside audio tracks.

1. The Processing Pipeline

When a user selects a Vamp plugin in Audacity, the host software routes the selected audio channel data to the plugin in framed blocks (often converted to the frequency domain using the Fast Fourier Transform, or FFT). The plugin processes these frames according to its designated algorithm and returns the extracted features back to Audacity as structured data points.

2. Output as Label Tracks

Rather than altering the waveform, Audacity plots the plugin's data output directly beneath the original audio using Label Tracks. For event-based or interval-based analyses—such as beat markers, note transcriptions, or structural segments—Audacity creates text labels synchronized precisely to the corresponding time in the audio waveform. This allows users to inspect exactly where specific musical or acoustic events occur.

3. Installation and Usage

To use Vamp plugins in Audacity:

Download compiled Vamp plugin files (which have extensions such as .dll on Windows, .so on Linux, or .dylib on macOS).
Move the files to the system's designated Vamp directory (such as C:\Program Files\Vamp Plugins on Windows or /Library/Audio/Plug-Ins/Vamp on macOS).
In Audacity, enable the plugin via the Plugin Manager.
Select an audio segment and run the plugin via the Analyze menu.

By converting raw waveforms into actionable metadata, Vamp plugins expand Audacity from a basic recording and editing suite into an effective tool for acoustic research and computational music analysis.