torchaudio_load_audio_spectrogram_transform_matplotlib_visualization.py

1import torch
2import torchaudio
3import torchaudio.transforms as T
4import matplotlib.pyplot as plt
5
6# Load an audio file (replace with your own file path)
7# torchaudio.utils.download_asset can be used to get sample data
8bundle = torchaudio.pipelines.WAV2VEC2_ASR_BASE_960H
9waveform, sample_rate = torchaudio.load("audio.wav") # Ensure audio.wav exists or use a sample
10
11# Define a transform (Spectrogram)
12spectrogram_transform = T.Spectrogram(n_fft=400)
13
14# Apply the transform to the waveform
15spectrogram = spectrogram_transform(waveform)
16
17# Print metadata
18print(f"Waveform shape: {waveform.shape}")
19print(f"Sample rate: {sample_rate}")
20print(f"Spectrogram shape: {spectrogram.shape}")
21
22# Optional: Visualize the spectrogram
23plt.figure()
24plt.imshow(spectrogram.log2()[0,:,:].numpy(), aspect='auto', origin='lower')
25plt.title("Spectrogram")
26plt.show()