Odd that it only hits the tick labels, but not the other text.

Can you please move the images and example code over here? I do not trust SO long-term.

Here is code to reproduce the pixelated fonts when saving a figure as a jpeg (when using jupyter):

%matplotlib nbagg

import matplotlib.pyplot as plt
import numpy as np

x    = np.linspace(-1.2,1.2,1000,endpoint=True)
y    = np.copy(x)
x,y  = np.meshgrid(x,y)
z    = -x**2 + y**2 - y**3

fig = plt.figure()
ax  = fig.add_subplot(111)

CS = plt.contour(x,y,z, [0,-0.1,0.1], colors=['black','blue', 'gray'])
plt.clabel(CS, fontsize=14, inline=1, fmt='%1.1f', manual=[(-0.15,0), (-0.4,0), (0.25,0.5)])

plt.title('Contour Sets')

plt.savefig('plot.png', format='png')
plt.savefig('plot.jpg', format='jpg')

This produces a plot.png shown here:

and plot.jpg shown here:

Adding the following code will "fix" this issue and produce jpegs that look nearly identical to the png files (with the undesirable side-effect of have non-transparent figures rendered in your jupyter notebooks):

import matplotlib
matplotlib.rcParams['nbagg.transparent'] = False

Another example for how PIL handles these transparent regions while trying to encode the raw data into jpeg is given by taking the first example and adding the code:

import matplotlib
matplotlib.rcParams['savefig.transparent'] = True

plot.jpg in this case looks like this:

I'm not exactly sure if this problem should be handled by the PIL team of the matplotlib team, but it might be worth looking into adding code to remove transparent regions when trying to save figures to file formats that can't handle transparency. I don't think it would be a difficult issue to fix, and might be able to write up a pull request for a suggested solution at some point in time.

This issue also seems to surface when trying to create animations.