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 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

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

# 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

# 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
)

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

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

Check out test_notebook.ipynb for a the visual output or download the example HTML file (demo.html).

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

2. Spectrogram Configuration

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

3. Waveform Configuration

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

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)
        }
    }
)