wavetable 0.1.2

A Python library to log audio files and their spectrograms into a single HTML file for easy inspection.

wavetable

A Python library for generating organized HTML tables of audio samples, waveforms, and spectrograms. wavetable allows you to compare "Generated" vs "Ground Truth" audio directly in your browser or Jupyter Notebook with zero-dependency static HTML output.

Features

• Static HTML Output: Generates self-contained (embedded) or lightweight (linked assets) HTML files.
• Visualizations: Automatic Waveform and Spectrogram (STFT or Mel) display.
• Comparisons: Native support for overlaying Ground Truth waveforms and displaying reference audio.
• Customizable: Control colormaps, frequency scales (Log/Mel/Linear), and magnitude scaling (dB/Linear).

Installation

pip install wavetable # if using pip
poetry add wavetable # if using poetry
...

Quick Start

import numpy as np
from wavetable import AudioLogger

# 1. Initialize
logger = AudioLogger(name="my_experiment",
                     sr=44100,
                     root_dir="audio_logs",
                     save_mode="embed" # or "link"
                    )

# 2. Create Dummy Audio
audio_gen = np.sin(2 * np.pi * 440 * np.linspace(0, 2, 44100*2))
audio_gt  = np.sin(2 * np.pi * 440 * np.linspace(0, 2, 44100*2)) + 0.1

# 3. Log a cell
logger.log(
    row="Epoch_1",
    col="Sample_A",
    audio=audio_gen,
    ground_truth=audio_gt,
    meta="Loss: 0.02"  # Text metadata displayed above the cell
)

# 4. Save to disk (creates audio_logs/my_experiment.html)
save_path = logger.save()

# Optional: Display in Jupyter Notebook
logger.display()

Configuration

You can customize the logger by passing a plot_config dictionary during initialization. The full configuration options available in the file wavetable/config.py.

1. General Settings

Parameter	Type	Default	Description
sr	int	44100	Global sampling rate.
save_mode	str	'embed'	'embed' (single huge HTML file) or 'link' (HTML + assets folder).
hover_effect	bool	True	Highlights the active row on mouse hover.
root_dir	str	'audio_logs'	Directory where logs are saved.

2. Spectrogram Configuration

Passed via plot_config={'spectrogram': {...}}:

Parameter	Options	Default	Description
spectrogram_type	'stft', 'mel'	'stft'	The type of spectrogram to compute.
plot_freq_scale	'log', 'linear'	'log'	Y-axis scaling (only used if type is 'stft').
magnitude_scale	'db', 'linear'	'db'	Compression of magnitude/power.
n_mels	int	80	Number of Mel bands (only used if type is 'mel').
cmap	str	'inferno'	Matplotlib colormap name (e.g., 'viridis', 'magma').
n_fft	int	2048	FFT window size.
hop_length	int	512	STFT hop length.

3. Waveform Configuration

Passed via plot_config={'waveform': {...}}:

Parameter	Type	Default	Description
overlay_gt	bool	False	If True, plots Ground Truth signal behind Generated signal.
color_gen	str	'#007bff'	Hex color for Generated audio (Blue).
color_gt	str	'#fd7e14'	Hex color for Ground Truth audio (Orange).
ylim	tuple	(-1, 1)	Y-axis limits for the waveform plot.

Advanced Example

logger = AudioLogger(
    name="Mel_Spectrogram_Test",
    sr=22050,
    hover_effect=False, # Disable row highlighting
    plot_config={
        'spectrogram': {
            'spectrogram_type': 'mel',
            'n_mels': 128,
            'magnitude_scale': 'db',
            'cmap': 'magma',
            'dimensions': (300, 80) # W, H in pixels
        },
        'waveform': {
            'overlay_gt': True,     # See Gen and GT on top of each other
            'dimensions': (300, 40)
        }
    }
)