memeplotlib 0.6.0

Memes with Python's matplotlib. Create image-macro memes by either letting the memegen API render them server-side (the default) or falling back to a local Pillow renderer when the API can't express what you want — custom local images, explicit per-line font sizes, custom outlines, or per-line position overrides.

Installation

pip install memeplotlib
# or
conda install -c conda-forge memeplotlib

For the Model Context Protocol server (use memes from Claude Desktop / Claude Code / any MCP client):

pip install "memeplotlib[mcp]"

Quick start

import memeplotlib as memes

fig, ax = memes.meme("buzz", "memes", "memes everywhere")
fig.savefig("buzz.png")

The function returns (Figure, Axes) — same convention as seaborn, pandas.plot, and other matplotlib extensions. It does not call plt.show() implicitly. Pass show=True if you want auto-display.

Features

Memegen IDs, file paths, or URLs as templates:

memes.meme("drake", "writing tests", "shipping to prod", color="yellow")
memes.meme("/path/to/image.jpg", "top text", "bottom text")
memes.meme("https://example.com/image.png", "from a URL")

Object-oriented Meme builder, chainable:

from memeplotlib import Meme

Meme("buzz").top("python").bottom("python everywhere").save("buzz.png")

Memify existing matplotlib figures:

import matplotlib.pyplot as plt

fig, ax = plt.subplots()
ax.plot([1, 2, 3], [1, 4, 9])
memes.memify(fig, "stonks")

RcParams-style scoped configuration:

with memes.rc_context({"font": "comic", "color": "yellow"}):
    memes.meme("buzz", "scoped style", "no leakage")
# defaults are auto-restored here

# Or set globally:
memes.config["fontsize"] = 96

Forward **kwargs to Axes.text:

memes.meme("buzz", "rotated", rotation=15, alpha=0.8)

Alt text for screen readers: every render gets a description built from the template and captions, embedded in saved PNG/SVG/PDF files:

fig, ax = memes.meme("buzz", "memes", "memes everywhere", savefig="buzz.png")
memes.get_alt_text(fig)
# 'Meme using the "Buzz Lightyear" template. Top text: "memes". Bottom text: "memes everywhere".'
memes.meme("buzz", "memes", alt_text="Buzz Lightyear pointing at memes everywhere")

Backends

memeplotlib ships three rendering backends. The default "auto" policy picks the best fit; pass backend="..." to override.

backend="auto" selects:

• memegen — when the template comes from the memegen catalogue and the caller didn't pass fontsize, a non-default outline_color / outline_width, **text_kwargs, or per-line overrides.
• matplotlib — when **text_kwargs (e.g. rotation, alpha) are passed, since only Axes.text can apply them.
• pillow — for custom local images / arbitrary URLs, or when any of the other client-only features above were requested.

# Default: memegen renders this server-side.
memes.meme("buzz", "memes", "memes everywhere", template_style="default",
           font="impact", width=600)

# Pillow fallback — explicit fontsize forces it.
memes.meme("/path/to/photo.jpg", "top", "bottom", fontsize=48,
           outline_color="red", outline_width=4)

# Per-line overrides force the Pillow backend.
from memeplotlib import Meme
Meme("buzz").top("hi").line(1, "world", fontsize=72, color="yellow").save("out.png")

# Axes.text kwargs select the matplotlib path automatically:
memes.meme("buzz", "rotated", rotation=15, alpha=0.8)

See docs/url_construction.rst for the full memegen URL grammar — escape table, query parameters, font / style / overlay reference — adapted from jacebrowning/memegen #993.

Use from agents

pip install "memeplotlib[mcp]"
memeplotlib-mcp                                 # boot the MCP server (stdio)
memeplotlib meme buzz "hello" "world" -o /tmp/hello.png  # CLI render

The MCP server exposes meme, search_templates, and list_templates tools. The CLI is useful even without MCP — agent harnesses can shell out, and CI scripts can render directly.

Documentation

Full docs including a tutorial, user guide, conventions reference, and API reference: brianckeegan.github.io/memeplotlib.

Build locally:

pip install -e ".[docs]"
sphinx-build -W docs docs/_build

How it works

1. Template metadata comes from the memegen API (/templates/, /templates/<id>); blank backgrounds and rendered memes alike are cached on disk.
2. Memegen backend (default for memegen IDs): a fully-formed rendering URL (/images/<id>/<line_1>/.../<line_n>.<ext>?style=...&font=...) is built via memeplotlib.build_memegen_url. The composed image is fetched and displayed with Axes.imshow.
3. Pillow backend (default for custom images, or when client-only features are requested): the blank is fetched once, then captions are drawn with PIL.ImageDraw using stroke-aware text rendering and a shrink-to-fit loop, and the composed RGBA array is imshown.
4. Matplotlib backend (legacy, opt-in): captions are drawn with Axes.text plus patheffects.Stroke for the classic outlined look.
5. The bundled Anton font (Impact-like, SIL OFL licensed) is used as a fallback for systems where Impact isn't installed.

Related projects

• matplotlib — the rendering engine.
• memegen — the template registry and blank-image source (api.memegen.link).
• seaborn — the API conventions for ax=None, **kwargs forwarding, and (fig, ax) returns are modeled on seaborn.

Dependencies

• matplotlib >= 3.8
• numpy, requests, Pillow, platformdirs

Requires Python 3.10+.

License

MIT. The bundled Anton font is licensed under the SIL Open Font License v1.1.