Matplotlib cannot find basic fonts

I am using matplotlib version 2.0.0 on Python 3 in a miniconda virtual environment. I am working on a unix scientific computing cluster where I don’t have root privileges. I am generally executing python code through an ipython notebook. If I do a basic command such as:

import matplotlib.pyplot as plt
plt.scatter([1,5], [1,5])

I get an error message:

packages/matplotlib/ UserWarning: findfont: Font family
['sans-serif'] not found. Falling back to DejaVu Sans
(prop.get_family(), self.defaultFamily[fontext]))

I would like to be able to use a Times New Roman font but even after deleting the Font cache file (fontList.py3k.cache) which I find from here:

import matplotlib as mpl
fm = mpl.font_manager

The commands:

mpl.rcParams[''] = ['serif']
mpl.rcParams['font.serif'] = ['Times New Roman']

have no effect, I get the same error as above. The true type fonts directory:


only has 40 fonts in it of the type: DejaVuSerif,DejaVuSans,STIX,cmb, cmt, cmy

Any idea what could be going on and how I can add additional fonts? Thanks!

Here is Solutions:

We have many solutions to this problem, But we recommend you to use the first solution because it is tested & true solution that will 100% work for you.

Solution 1

To get it to work, I had to combine the two current top answers. Here’s what worked for me:

$ sudo apt install msttcorefonts -qq
$ rm ~/.cache/matplotlib -rf

Solution 2

I had this exact same problem on a Vagrant VM running Ubuntu Xenial 64-bit. No matter how many fonts I had already installed, matplotlib was having a problem with the “system” font name “sans-serif”. I fixed it by:

  • Stopping Jupyter
  • Installing font-manager: sudo apt install font-manager
  • Cleaning the matplotlib cache directory: rm ~/.cache/matplotlib -fr
  • Restarting Jupyter.

No more error messages about sans-serif.

Solution 3

This work for me::

$ sudo apt-get install msttcorefonts -qq

Solution 4

A solution for Windows users, when confronted with the warning:

UserWarning: findfont: Font family ['serif'] not found. Falling back to DejaVu Sans
(prop.get_family(), self.defaultFamily[fontext]))
  1. Delete the fonts located in matplotlib’s cache.
    Cache’s location: import matplotlib as mpl; print(mpl.font_manager.get_cachedir())

  2. Find matplotlib’s font directory. The path might be similar to

  3. Copy necessary fonts like Computer Modern to this directory.

The warning may persist, but your plots’ font should change appropriately.

Solution 5

There is a conda package for it[1].
So, you don’t really need sudo to fix this!

conda install -c conda-forge -y mscorefonts

UPDATE: the below functionality is broken in recent matplotlib (no longer works on v3.4.3)

Also, if you dont want to restart jupyter, you can force rebuild matplotlib font cache

import matplotlib
[1] –

Solution 6

I actually solved by

import matplotlib.font_manager

Solution 7

I had this problem with anaconda env (it could be useful in other related situations as well).
For example, for base env, I have checked the following directory:

and noticed that I have DejaVuSerif.ttf intalled there, so instead of using plt.rcParams[‘’] = ‘Serif’ I used plt.rcParams[‘’] = ‘DeJavu Serif’ ( there is a space between "DeJavu and serif" ). i worked for me.

import matplotlib.pyplot as plt
plt.rcParams[''] = 'DeJavu Serif'
plt.rcParams['font.serif'] = ['Times New Roman']

Solution 8

It took me many hours to finally figure out that I needed to update matplotlib.

My original version of 3.0.3 (which came with my jupyter-datascience docker image) would give me this error, but updating to 3.1.1 fixed it.

In the end, my docker script:

RUN pip install matplotlib==3.1.1
COPY fonts /usr/share/fonts/truetype/
RUN fc-cache -fv
RUN rm /home/jovyan/.cache/matplotlib -rf

is what did it for me.

Solution 9

I was facing a similar issue in a Cloud Datalab docker image running on a gcloud VM. Executing the following fixed the issue for me:

$ sudo apt install msttcorefonts -qq
$ rm ~/.cache/matplotlib -rf

Here is instructions on how to get to the docker image running on the gcloud VM containing the Datalab instance just in case.

Solution 10

try all the methods above, not work for me.

my way to solve this is a little dirrent cause I’m using conda on ubuntu server running jupyter

locate -b '\mpl-data'

find a folder


then I add the simhei fonts into the font folder in mpl-data .
then remove matplotlib fonts cache

rm -rf ~/.cache/matplotlib

restart jupyter notebook, and it works.

Solution 11

I am in macOS with jupyter notebook, I solved with the following, first close your jupyter notebook. Then find out the font path by doing the following in Python

import matplotlib

it prints /Users/zyy/anaconda2/lib/python2.7/site-packages/matplotlib/mpl-data/matplotlibrc for me, notice matplotlibrc is a file, not a directory.

Then download font SimHei, and copy it to the directory fonts/ttf under the mpl-data/ directory above.

Delete directory ~/.cache/matplotlib and restart your jupyter notebook, everything should be good.

Solution 12

For windows users

  1. Just go to the cache dir of matplotlib by using

import matplotlib as mpl


  1. Clear the entire cache file

  2. Lastly again import matplotlib

Hopefully u will also find it helpful as it worked for me

P:S – If you wish you can restart your jupyter/ide (optional) after this

Solution 13

None of the above worked for me. I had to combine the provided solutions.

I run Jupyter from PyCharm in Windows 10. Here is my workaround. Before doing these steps, import the font manager via import matplotlib.font_manager:

  1. Find font directory => matplotlib.font_manager.findSystemFonts(fontpaths=None, fontext='ttf')
  2. Put your font file (ttf. file) to that directory.
  3. Find font config file: matplotlib.matplotlib_fname()
  4. Edit to set font type: pdf.fonttype: 42 # Output Type 3 (Type3) or Type 42 (TrueType)
  5. Run PyCharm.
  6. After you start kernel, cache files will be generated. Clean \matplotlib\__pycache__ folder.
  7. Set the font you want to use via:
plt.rcParams[''] = 'serif'
plt.rcParams['font.serif'] = ['Lin Libertine']
  1. Run the code to generate plots.

If you restart your kernel, if will generate cache files, hence you need to remove them again.

Solution 14

My situation is that I have JupyterHub running in Docker and I need to use matplotlib to plot and display some Chinese characters as titles.

Here’s how I install new fonts and get new fonts worked( No need to restart the Jupyter container ):

Docker Image version:

  • Ubuntu 18.04.3 LTS
  • Python 3.7.6
  • matplotlib.__version__ : 3.1.2

All operations are executed inside the container.

  1. download the fonts to /.../python3.7/site-packages/matplotlib/mpl-data/fonts/ttf

  2. rebuild the font cache in a notebook. ~/.cache/matplotlib/ (AttributeError: module 'matplotlib' has no attribute 'font_manager').

import matplotlib.font_manager

the font cache is per-user base(/home/{username}/.cache/matplotlib/), so other people might need to rebuild the cache too.

  1. use the new font. e.g.
import matplotlib.pyplot as plt

plt.rcParams['font.sans-serif'] = ['SimSun']

You might need to restart the notebook/kernel in step 2 and 3.


  • I tried sudo apt install msttcorefonts -qq but I don’t think that works. Neither do I know what has been installed by that package. apt-file list msttcorefonts returns nothing.
    Why apt-file list msttcorefonts returns nothing?

  • There is a configuration file /.../python3.7/site-packages/matplotlib/mpl-data/matplotlibrc on which one can add font.sans-serif fonts. But without modifying it, the new fonts also work.

Note: Use and implement solution 1 because this method fully tested our system.
Thank you 🙂

All methods was sourced from or, is licensed under cc by-sa 2.5, cc by-sa 3.0 and cc by-sa 4.0

Leave a Reply