kikitian / celluloid

:movie_camera: Matplotlib animations made easy

celluloid

Easy Matplotlib Animation

Creating animations should be easy. This module makes it easy to adapt your existing visualization code to create an animation.

Install

pip install celluloid

Manual

Follow these steps:

1. Create a matplotlib Figure and create a Camera from it:

from celluloid import Camera
fig = plt.figure()
camera = Camera(fig)

2. Reusing the figure and after each frame is created, take a snapshot with the camera.

plt.plot(...)
plt.fancy_stuff()
camera.snap()

3. After all frames have been captured, create the animation.

animation = camera.animate()
animation.save('animation.mp4')

The entire module is less than 50 lines of code.

Viewing in Jupyter Notebooks

View videos in notebooks with IPython.

from IPython.display import HTML
animation = camera.animate()
HTML(animation.to_html5_video())

Examples

Minimal

As simple as it gets.

from matplotlib import pyplot as plt
from celluloid import Camera

fig = plt.figure()
camera = Camera(fig)
for i in range(10):
    plt.plot([i] * 10)
    camera.snap()
animation = camera.animate()

Subplots

Animation at the top.

import numpy as np
from matplotlib import pyplot as plt
from celluloid import Camera

fig, axes = plt.subplots(2)
camera = Camera(fig)
t = np.linspace(0, 2 * np.pi, 128, endpoint=False)
for i in t:
    axes[0].plot(t, np.sin(t + i), color='blue')
    axes[1].plot(t, np.sin(t - i), color='blue')
    camera.snap()
animation = camera.animate()

Images

Domain coloring example.

import numpy as np
from matplotlib import pyplot as plt
from matplotlib.colors import hsv_to_rgb

from celluloid import Camera

fig = plt.figure()
camera = Camera(fig)

for a in np.linspace(0, 2 * np.pi, 30, endpoint=False):
    x = np.linspace(-3, 3, 800)
    X, Y = np.meshgrid(x, x)
    x = X + 1j * Y
    y = (x ** 2 - 2.5) * (x - 2.5 * 1j) * (x + 2.5 * 1j) \
        * (x - 2 - 1j) ** 2 / ((x - np.exp(1j * a)) ** 2
        * (x - np.exp(1j * 2 * a)) ** 2)

    H = np.angle(y) / (2 * np.pi) + .5
    r = np.log2(1. + np.abs(y))
    S = (1. + np.abs(np.sin(2. * np.pi * r))) / 2.
    V = (1. + np.abs(np.cos(2. * np.pi * r))) / 2.

    rgb = hsv_to_rgb(np.dstack((H, S, V)))
    ax.imshow(rgb)
    camera.snap()
animation = camera.animate()

Legends

import matplotlib
from matplotlib import pyplot as plt
from celluloid import Camera

fig = plt.figure()
camera = Camera(fig)
for i in range(5):
    t = plt.plot(range(i, i + 5))
    plt.legend(t, [f'line {i}'])
    camera.snap()
animation = camera.animate()

Limitations

• The axes' limits should be the same for all plots. The limits of the animation will be the limits of the final plot.
• Legends will accumulate from previous frames. Pass the artists to the legend function to draw them separately.
• Animating the title does not work. As a workaround you can create a text object:

ax.text(0.5, 1.01, 'computed title', transform=ax.transAxes)

• This can demand a lot of memory since it uses ArtistAnimation under the hood. This means that all artists are saved to memory before the animation is constructed.
• This is a black box. If you want to understand how matplotlib animations work, using this library may hinder you. If you want to be an expert matplotlib user, you may want to pass on this library.

Credits

Inspired by plotnine.