Bring your audio
Read files, capture a microphone, or generate a signal. Work with one consistent AudioTensor representation.
WAV · FLAC · OGG · optional MP3 Python library v0.2.2
Audio processing.
Expressed in Python.
From the first sample to the final feature.
Record, transform, and analyze sound with a consistent tensor API.
python -m pip install phonotensor A signal. A tensor. A possibility.01 — 03
01 / INPUT
Sound comes in.
audio.shape → (88200, 1)02 / TRANSFORM
Ideas take shape.
ops.lowpass(audio, 2000)ops.normalize(audio, 0.8)ops.stft(audio)
samples → frames × bins03 / ANALYZE
Patterns emerge.
FeatureTensor → NumPyAn illustration of the workflow No audio is recorded in your browser
A SMALLER GAP BETWEEN IDEA AND SIGNAL
One vocabulary for your audio.
Keep sample rates, channels, and features in view as your workflow grows.
Shape the signal
Compose filters, fades, edits, mixing, and resampling with ordinary Python functions.
NumPy arrays · explicit sample ratesFind the features
Move from samples to spectra, mel features, and MFCCs, with frame metadata carried alongside the data.
STFT · mel spectrograms · MFCCLESS SETUP. MORE EXPLORING.Your first signal,
Your first signal,
in a few lines.
Generate a tone, soften its edges, and save a file. Start without a microphone, then bring in your own audio.
Follow the quickstartfrom phonotensor import FileIO, ops
# Create a two-second tone. No microphone needed.
audio = ops.sine(440, duration=2.0, amplitude=0.25)
audio = ops.fade_in(audio, duration=0.02)
audio = ops.fade_out(audio, duration=0.1)
FileIO.write(audio, "tone.wav")
print(audio.shape) # (88200, 1)
print(audio.sample_rate) # 44100Make something worth hearing.
Installation guides, practical examples, and the API—ready when you are.
Read the documentation