Matplotlib Colors and Colormaps: Named Colors, Hex Codes, cmaps

How to set a color in Matplotlib

Any function that draws something takes a color argument (c for short in plot and scatter). Matplotlib understands all of these:

Format	Example	Notes
Named color	`color="tomato"`	Any of the 148 CSS names below.
Hex code	`color="#1f77b4"`	Add two digits for transparency: `"#1f77b480"`.
Tableau color	`color="tab:orange"`	The ten colors of the default cycle.
Cycle color	`color="C2"`	The third color of the current cycle (C0 is the first).
One letter	`color="r"` or `"r--"`	`b g r c m y k w`; also works in format strings.
RGB or RGBA tuple	`color=(0.1, 0.4, 0.8, 0.5)`	Values from 0 to 1, not 0 to 255.
Gray level	`color="0.75"`	A string from `"0"` (black) to `"1"` (white).

colors.py

Run this chart →

colors.py

import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(0, 10, 100)
fig, ax = plt.subplots(figsize=(8, 5))

# Six ways to give a color
ax.plot(x, np.sin(x), color="tomato", label='"tomato" (named color)')
ax.plot(x, np.sin(x) + 1, color="#2a9d8f", label='"#2a9d8f" (hex code)')
ax.plot(x, np.sin(x) + 2, color="tab:purple", label='"tab:purple" (Tableau)')
ax.plot(x, np.sin(x) + 3, color=(0.9, 0.6, 0.1), label="(0.9, 0.6, 0.1) (RGB tuple)")
ax.plot(x, np.sin(x) + 4, color="C0", label='"C0" (first cycle color)')
ax.plot(x, np.sin(x) + 5, color="0.4", label='"0.4" (gray level)')

ax.legend(loc="upper right", fontsize=8)
ax.set_title("Ways to set a color in Matplotlib")
plt.show()

Default color cycle (Tableau colors)

When you do not choose a color, each new line or bar set takes the next of these ten colors. Use them by name ("tab:blue") or by position ("C0" to "C9").

Single-letter colors

Named colors (CSS4)

All 148 names Matplotlib accepts, grouped by hue. Select one to copy its name. "grey" and "gray" spellings both work.

Matplotlib colormaps (cmap)

A colormap turns numbers into colors in scatter, imshow, pcolormesh, contourf and heatmaps. Pass its name as cmap="viridis", and add _r for the reversed map ("viridis_r"). Select a colormap to copy its name.

Perceptually uniform sequential

Sequential

Diverging

Cyclic

Qualitative

Miscellaneous

colormaps.py

Run this chart →

colormaps.py

import matplotlib.pyplot as plt
import numpy as np

rng = np.random.default_rng(1)
x, y = rng.normal(size=(2, 400))
distance = np.hypot(x, y)

fig, (left, right) = plt.subplots(1, 2, figsize=(10, 4))

# A colormap turns numbers into colors; the colorbar is its legend.
points = left.scatter(x, y, c=distance, cmap="viridis", s=18)
fig.colorbar(points, ax=left, label="distance from center")
left.set_title('cmap="viridis"')

# A diverging map for data above and below zero
grid = np.sin(np.linspace(-3, 3, 60))[:, None] * np.cos(np.linspace(-3, 3, 60))[None, :]
image = right.imshow(grid, cmap="coolwarm", vmin=-1, vmax=1)
fig.colorbar(image, ax=right)
right.set_title('cmap="coolwarm"')

plt.tight_layout()
plt.show()

Which colormap to choose

Low to high values: sequentialviridis, cividis, plasma or a single hue such as Blues. Lightness grows evenly, so equal steps in the data look like equal steps in color, also when printed in gray.
Above and below a midpoint: divergingcoolwarm, RdBu, PiYG. Set vmin and vmax to the same distance from the midpoint, for example -1 and 1, so zero falls on the neutral color.
Categories: qualitativetab10, Set2, Dark2. Distinct colors with no order. Use them for groups, not for numbers.
Angles and time of day: cyclictwilight and hsv start and end on the same color.
Avoid jet and rainbowBright bands make some changes look larger than they are and hide others, and they are hard to read for color-blind viewers. turbo is a smoother alternative if you need many hues.

Get colors from a colormap

Call a colormap with numbers from 0 to 1 to get RGBA colors, for example to give several lines related colors.

pick_colors.py

Run this chart →

pick_colors.py

import matplotlib
import matplotlib.pyplot as plt
import numpy as np

# Five evenly spaced colors from a colormap, one per line
colors = matplotlib.colormaps["plasma"](np.linspace(0, 0.9, 5))

x = np.linspace(0, 2 * np.pi, 200)
for i, color in enumerate(colors):
    plt.plot(x, np.sin(x + i * 0.5), color=color, linewidth=2, label=f"line {i + 1}")

plt.legend()
plt.title("Colors taken from the plasma colormap")
plt.show()

Change the default colors

Set the color cycle once, and every plot after it uses your colors in order.

color_cycle.py

Run this chart →

color_cycle.py

import matplotlib.pyplot as plt
import numpy as np

# Every new line takes the next color of the cycle
plt.rcParams["axes.prop_cycle"] = plt.cycler(color=["#e63946", "#457b9d", "#2a9d8f", "#f4a261"])

x = np.linspace(0, 10, 100)
for shift in range(4):
    plt.plot(x, np.sin(x) + shift, linewidth=2)

plt.title("A custom default color cycle")
plt.show()

Matplotlib lessons

Frequently asked questions

How do I list all named colors in Matplotlib?

matplotlib.colors.CSS4_COLORS is a dictionary of the 148 CSS color names and their hex codes. BASE_COLORS has the one-letter colors and TABLEAU_COLORS the tab: colors of the default cycle.

How do I reverse a colormap?

Add _r to its name, for example cmap="viridis_r". Every built-in colormap has a reversed version.

How do I get a list of colors from a colormap?

Call the colormap with numbers from 0 to 1: matplotlib.colormaps["viridis"](np.linspace(0, 1, 5)) returns five RGBA colors. resampled(5) returns a colormap with five colors.

How do I change the default colors of lines?

Set the color cycle: plt.rcParams["axes.prop_cycle"] = plt.cycler(color=["#e63946", "#457b9d", "#2a9d8f"]). New plots then use those colors in order.

Which colormap should I use?

Use a perceptually uniform map such as viridis or cividis for values that go from low to high, a diverging map such as coolwarm or RdBu when values go above and below a midpoint, and tab10 for categories. Avoid jet and rainbow, which make some changes look larger than they are.

How do I make a color transparent?

Pass alpha, for example plt.plot(x, y, color="tab:blue", alpha=0.4), or use an RGBA tuple or an 8-digit hex code such as "#1f77b466".