Saving Figures

savefig is the output side of every chart: raster (PNG, JPG) for screens and messaging, vector (SVG, PDF, EPS) for publications and infinite zoom. The method writes the Figure exactly as laid out - which makes two parameters worth knowing by heart: bbox_inches="tight" to clip the margin around the actual content, and dpi to set raster resolution. Everything else is format-specific detail.

Basic Saving

import matplotlib.pyplot as plt

x = [0, 1, 2, 3, 4]
fig, ax = plt.subplots()
ax.plot(x, [v * v for v in x], marker="o")

fig.savefig("parabola.png")            # raster, default 100 dpi
fig.savefig("parabola.svg")             # vector
fig.savefig("parabola.pdf")             # vector
plt.show()

The format comes from the file extension. The figure is saved in its current state - save before adding final annotations if you want the clean version, or just save the finished state. savefig does not close the figure; pair it with plt.close(fig) in loops and server code to free memory.

Tight Bounds and Resolution

fig.savefig("chart.png", dpi=200, bbox_inches="tight")

• bbox_inches="tight" recomputes the saved bounding box to the content, dropping the default figure margins - the single most useful save parameter, because it fixes clipped legends and titles.
• dpi: screen targets want 100-150; print and zoom targets 200-300. A 6x4-inch figure at 200 dpi is a 1200x800 pixel image.
• facecolor/edgecolor set the canvas color; transparent=True saves PNG/SVG with no background - the standard for logos and overlay graphics.

Raster vs Vector

Vector formats store geometry, not pixels: text stays crisp at any size and file size scales with complexity, not resolution. If the figure goes in a paper, a thesis, or a slide deck that gets zoomed, save SVG or PDF. If it goes in a dashboard or a message, PNG at 150+ dpi is fine.

Saving in Loops and Services

import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt

for i in range(5):
    fig, ax = plt.subplots()
    ax.plot([0, 1, 2], [i, i + 1, i])
    fig.savefig(f"series_{i}.svg", bbox_inches="tight")
    plt.close(fig)        # release the canvas - critical in long loops

The pattern: create, draw, save, close. In a web service that renders charts on demand, the Agg backend plus close is the entire rendering story - no display, no window, no accumulated memory. For very frequent small renders, fig.canvas.print_figure writes to a buffer you can stream directly to a response.

Summary

Save with the extension choosing the format, bbox_inches="tight" to fix margins, and dpi set for the destination. PNG for screens, SVG/PDF for anything printed or zoomed, transparent backgrounds for overlays, and plt.close after every save in non-interactive code. Ten seconds of parameter discipline and the output stops being an afterthought.