Skip to content

Jupyter and JupyterLab

This page changed after phonometry 4.0.0, the latest release on PyPI, so it may describe code that release does not have yet. What is new since then is listed under Unreleased in the changelog.

A notebook runs phonometry in an ordinary IPython kernel, so it installs as anywhere else; what the notebook adds is how a result shows itself under a cell. In a notebook, in Getting Started, walks through a first cell and the table it displays; this page is the rest of what a notebook user meets.

Checked on 17 November 2026 with phonometry 4.0.0 from PyPI, in the python:3.13-slim image on Python 3.13.18: JupyterLab 4.6.6, ipykernel 7.4.0, matplotlib 3.11.2 and Debian’s ffmpeg 7.1.5, with every cell of this page run in the kernel.

Terminal window
python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install jupyterlab "phonometry[full]"
jupyter lab

On Windows the environment is activated with .venv\Scripts\activate.bat in cmd or .venv\Scripts\Activate.ps1 in PowerShell in place of source .venv/bin/activate. PowerShell on a Windows client refuses scripts under its default execution policy, so Activate.ps1 runs there once Set-ExecutionPolicy -Scope CurrentUser RemoteSigned has allowed local ones; cmd needs no such step.

That one environment holds JupyterLab, the IPython kernel it starts and phonometry with every extra but [torch].

On 32-bit Windows and 32-bit Linux numba is not supported, and neither is [full], which holds it: python -m pip install jupyterlab "phonometry[plot,report,audio]" installs JupyterLab with the other three extras. PyPI has no 32-bit Windows wheel of SciPy, matplotlib or kiwisolver, nor of psutil and pywinpty, which JupyterLab needs there, nor of PyYAML on Python 3.14, so pip builds those from source there, which takes a C and a C++ compiler, a Fortran compiler with BLAS and LAPACK for SciPy, and for pywinpty Rust and the NuGet command-line tool, nuget.exe, on PATH. On 32-bit Linux, measured on the 32-bit images of python:3.13-slim, PyPI has no wheel of NumPy, SciPy, matplotlib, Pillow, contourpy, kiwisolver, psutil, PyYAML, tornado or argon2-cffi-bindings, nor of cffi and pyzmq on ARM, of MarkupSafe on linux/386 and linux/arm/v5, or of lxml and rpds-py on linux/arm/v5, so pip builds those there, with the same compilers, the development files of the libraries they link and Rust for rpds-py. On linux/arm/v5, whose soft-float ABI no manylinux wheel targets, the builds of NumPy, SciPy, matplotlib and pyzmq also ask for ninja, patchelf and cmake, and pip would build those from source too, which runs make and autoreconf, unless apt-get install ninja-build patchelf cmake has put the image’s own on PATH.

The value on the last line of a cell displays as a table of its fields, with their units, and the line under it names what the result can draw:

import numpy as np
from phonometry import building, filters, io
fs = 48000
t = np.arange(fs) / fs
x = 0.4 * np.sin(2 * np.pi * 1000.0 * t)
bands = filters.octave_filter(x, fs, fraction=3)
bands # the last line of the cell: the notebook shows its table

.plot() draws under the cell through the inline backend, with no plt.show(). The cell also prints the axes it returns, as <Axes: title={'center': 'Band levels'}, ...>; a semicolon after the call leaves the figure alone. The figures come out as PNG; one line in a cell switches the session to SVG:

%config InlineBackend.figure_formats = ["svg"]
bands.plot();

A .plot() that raises, on a misspelt keyword such as colour= for instance, leaves no half-drawn figure under the cell, only the error.

rating = building.weighted_rating(
[30, 34, 38, 41, 45, 49, 50, 53, 54, 55, 56, 57, 58, 58, 58, 58]
)
rating.report("rating.pdf")

The PDF is written to the path you give, relative to the notebook’s working directory; [report] writes it and [plot] draws its figure.

io.write_video and the FDTD animation writers run ffmpeg, a program that pip does not install. On Debian it comes from the system’s packages, with apt install ffmpeg, and Writing videos gives the command for conda, macOS and Windows. IPython’s own display classes then show the file under the cell:

from matplotlib.animation import FuncAnimation
from matplotlib.figure import Figure
fig = Figure(figsize=(3.2, 2.4), dpi=100)
ax = fig.add_subplot()
phase = np.linspace(0.0, 2.0 * np.pi, 200)
(line,) = ax.plot(phase, np.sin(phase))
def update(k):
line.set_ydata(np.sin(phase - 0.2 * k))
return (line,)
animation = FuncAnimation(fig, update, frames=30, interval=50)
io.write_video("wave.mp4", animation)
io.write_video("wave.gif", animation, video=io.VideoOptions(quality="small"))
from IPython.display import Image, Video, display
display(Video("wave.mp4", embed=True)) # an MP4 or a WebM, embedded in the notebook
display(Image(filename="wave.gif")) # a GIF

A cell displays the value of its last line only, so both go through display: as two bare lines, the cell would show the GIF and not the video.

A warning the library gives names the line of the cell that made the call, and it shows under every cell that makes one, not only the first:

/tmp/ipykernel_2485/1280619122.py:1: FilterBankWarning: Low sampling rate: fs = 48000 Hz cannot hold the 25 kHz band, whose upper edge 28184 Hz is above fs/2 = 24000 Hz; it was left out of the bank.
wide = filters.octave_filter(x, fs, fraction=3, limits=[12.0, 25000.0])

To have warnings raise instead, set the filter in a cell. All of them:

import warnings
warnings.simplefilter("error")

or the library’s alone, warnings.simplefilter("error", phonometry.PhonometryWarning). Do not start the kernel with PYTHONWARNINGS=error: the kernel itself then stops on a DeprecationWarning of its own event loop before it answers the notebook, and the notebook reports a kernel that died.

Both run notebooks on an IPython kernel, so what this page shows should apply there as well. Neither was tried for this page.