meme-image

name

meme-image

description

Generate image memes via the memegen.link API for embedding in blog posts, essays, and other markdown contexts. The API is stateless, free, no auth required — every meme is fully specified in its URL. Use this skill when the user asks for an image meme, a galaxy-brain diagram, a Drake post, an expanding-brain comparison, a Distracted Boyfriend, or any classic meme template populated with custom text. Also use when adding visual rhythm to a long post — the screen heuristic below codifies how often. Do not use this for SVG illustrations, diagrams, or visual explainers — those are different (frontend-design, or the SVG guidance in franklin-essay, a separate optional skill that may not be installed). This skill is specifically for the recognizable meme templates that carry humor in their format.

meme-image

Generate image memes via memegen.link, a free, stateless, public API where the URL is the complete specification of the meme. No auth, no signup, no rate limits to worry about for normal blog use. The output is a PNG (or JPG, GIF, WebP) that you embed in markdown with a standard ![alt](url) tag.

When to use this skill

Use when the user asks for:

• a meme image to put in a blog post or essay
• a "galaxy brain" / "expanding brain" comparison
• a "drake" yes/no comparison
• "two buttons" choice meme
• "distracted boyfriend"
• "change my mind"
• "this is fine"
• any other recognizable meme template populated with their custom text
• a quick visual joke that needs the cultural shorthand of a known template
• visual rhythm in a long post (see screen heuristic below)

Do NOT use this skill when:

• the user wants an original SVG illustration (use frontend-design or inline SVG)
• the user wants a diagram or flowchart (use mermaid)
• the user wants an actual photograph (use image search or web fetch)
• the visual carries a thesis the post can't afford to be light about (legal documents, condolences, medical writing, anything marked as serious)

The screen heuristic

A reader on a phone sees about 400–600 words of body text per scroll, depending on font size and line-height. On gwern-typography (Franklin's blog) it's closer to 400 because the leading is generous. The pulse of a well-designed long post follows this rhythm:

Every 400–500 words, the reader meets a visual element to rest on.

That element doesn't have to be a meme. It can be:

• an image meme (this skill)
• a text meme block (greentext, dialogic, escalation list — see text-meme-injection skill)
• an SVG illustration
• a mermaid diagram
• an embedded map
• a footnote that pays off when clicked
• a pull quote
• a section header

But it has to be something — a moment where the reading eye gets a break before the next paragraph block. Mobile is the binding constraint — desktop fits more words per screen, but design for the tighter reader. Posts that violate this rhythm (walls of prose for 800+ words) feel enfadonhos on mobile no matter how good the writing. Posts that overdo it (visual every 100 words without breathing room) feel like content marketing.

For a 2,000-word post, the 400–500-word cadence means ~4–5 visual rests; for 2,500 words, ~5–6. This is the same cadence the text-meme-injection skill uses for inline memes in non-serious registers — the two skills share one rhythm. Image memes are heavier than other rest types and should account for roughly 1/4 to 1/3 of the rests in a meme-friendly register; the remaining rests are text memes, diagrams, SVGs, maps, footnotes, pull quotes, and section headers.

Default register

Unless the user explicitly marks a post as serious, assume non-serious register. Most posts on Franklin's blog are philosophical-but-irreverent, deadpan-Borges-coded — exactly the register where memes belong as native vocabulary, not as decoration.

This means:

• Multiple image memes per post are expected, not exceptional.
• Image memes can sit anywhere a visual rest is needed, not only at openings or as set pieces.
• The reflexive landing rule still applies but is gentler — a single hedging sentence after a thesis-bearing meme is enough; the meme doesn't need a full exoneration paragraph. Memes that don't carry a contestable thesis (pure tonal pieces) need no reflexive landing at all.
• No two image memes within ~500 words of each other — they need breathing room to land separately.

The serious-register rules (one image meme max, only at openings, full reflexive landing) apply only when the user explicitly says "this post is serious" or the topic is legal, grief, illness, abuse, addiction, suicide, medical, or condolences.

How memegen.link URLs work

The URL is the spec. Format:

https://api.memegen.link/images/{template_id}/{line1}/{line2}/.../{lineN}.{ext}

• template_id — short ID from the templates list (e.g., gb for galaxy brain, drake, cmm for change my mind, pooh, db for distracted boyfriend, fine for this is fine, bus for two guys on a bus)
• lineN — text for each text slot in the template, in order
• ext — png (default), jpg (smaller), gif or webp (animated if template supports)

Encoding rules

In URL path text:

• space → underscore _ or dash -
• literal underscore → __
• literal dash → --
• newline → ~n
• question mark → ~q
• apostrophe → ~s
• quotation → ~d
• percent → ~p
• hash → ~h

Avoid spaces, slashes, and exotic punctuation when you can. Most blog memes work with plain ASCII alphanumeric plus underscores.

Query parameters

Append ?param=value&param2=value2 to customize:

• width=600 — set width in pixels
• height=600 — set height; if both width and height set, image is padded
• font=impact / font=titilliumweb / font=notosans — change font
• layout=top — for some templates, force text to top instead of split
• style=<name> — for templates with multiple styles (rare, but exists for ds, pigeon, etc.)

Examples

Galaxy brain, four lines:

https://api.memegen.link/images/gb/Know_thyself/Nothing_in_excess/E/If_you_want_to_talk_to_the_Oracle_go_to_Delphi.png?width=600

Drake, two lines:

https://api.memegen.link/images/drake/Cogito_ergo_sum/Loxias_says_cross_the_Halys.png?width=500

Two guys on a bus, two lines:

https://api.memegen.link/images/bus/My_persona/My_persona_after_self-disclosing.png?width=600

This is fine, two lines:

https://api.memegen.link/images/fine/2500_years_uptime/no_plans_to_ship_a_fix.png?width=500

Change my mind, single line:

https://api.memegen.link/images/cmm/Sócrates_was_the_first_red-teamer.png?width=500

Listing available templates dynamically

The memegen.link template catalog grows. New templates appear; old ones occasionally get reorganized. Hardcoded lists go stale. Always read the full live catalog before choosing a template — every time, not only when you're unsure one exists. The shortlist reference file is a memory aid, not the menu, and memory is exactly where template-default bias lives: the reflex reaches for the half-dozen templates that come to mind, and "comes to mind" is the bias, not a fit signal. Reading the full list first puts the non-obvious templates back in play.

Full list (a couple hundred templates; always read the live catalog for the current list):

curl -s https://api.memegen.link/templates/ | python3 -c "
import json, sys
data = json.load(sys.stdin)
print(f'Total: {len(data)}')
for t in data:
    print(f\"  {t['id']:25s} | lines={t.get('lines','?')} | {t.get('name','')}\")
"

Filtered by keyword (e.g., looking for politicians, animals, brain memes):

KEYWORD="brain"
curl -s https://api.memegen.link/templates/ | python3 -c "
import json, sys, os
data = json.load(sys.stdin)
keyword = os.environ.get('KEYWORD', '').lower()
for t in data:
    name = t.get('name', '').lower()
    tid = t.get('id', '').lower()
    if keyword in name or keyword in tid:
        print(f\"  {t['id']:25s} | lines={t.get('lines','?')} | {t.get('name','')}\")
"

Each template's full spec (lines, styles, example):

curl -s "https://api.memegen.link/templates/gb"

When the user mentions a meme by colloquial name and the obvious ID doesn't match (e.g., they say "expanding brain" → try gb; they say "galaxy brain" → also gb; they say "drake hotline bling" → try drake), the dynamic catalog is the source of truth, not memory.

Draw against your defaults

Reading the catalog is necessary but not sufficient. Even with the full couple-hundred-template list in front of you, the hand still drifts to the marquee templates (drake, gb, tb, db, fine, cmm) — and a post whose central move is a reject/prefer binary will pull drake like a magnet, which is exactly when the obvious choice is least examined. So after reading the catalog and before settling on a template, draw a random handful and look at them. The draw is mandatory to perform. It is never mandatory to use.

curl -s https://api.memegen.link/templates/ | python3 -c "
import json, sys, random
d = json.load(sys.stdin)
for t in random.sample(d, 12):
    print(f\"  {t['id']:18s} | lines={t.get('lines','?')} | {t.get('name','')}\")
"

For each drawn template ask one question only: does it carry a real beat in this specific post? Most won't — a draw of 12 might yield two fits or zero, and that is the expected outcome, not a failed draw. Never use a template just because it came up; function-first still governs, so if the beat is a clean comparison and drake carries it best, drake wins even against everything the draw surfaced. When presenting options, report the draw honestly — say which drawn templates fit and which didn't; the discards show the search was real.

The payoff is debiasing, not randomness. A draw that ends with the obvious template is a success if you confirmed it beat the alternatives; a draw that surfaces a better non-obvious fit — rollsafe over drake for a "found the loophole" beat — is the case the rule exists for.

Template shortlist (reference file)

A non-exhaustive shortlist of known-good templates for essay-style posts — grouped by function (comparison and duality, escalation, single thesis, reaction, choice, stonks) — lives in references/shortlists.md, relative to this skill's directory. It is a memory aid, not the menu: always treat the live API catalog as canonical. For specific cultural references not in the shortlist, run the keyword-filtered catalog command above.

How to embed in markdown

Standard markdown image syntax works:

![Galaxy brain of the three Delphic inscriptions plus the Borland joke](https://api.memegen.link/images/gb/Know_thyself/Nothing_in_excess/E/If_you_want_to_talk_to_the_Oracle_go_to_Delphi.png?width=600)

For better presentation in essay-style posts, wrap in a <figure>:

<figure class="meme">
  <img
    src="https://api.memegen.link/images/gb/Know_thyself/Nothing_in_excess/E/If_you_want_to_talk_to_the_Oracle_go_to_Delphi.png?width=600"
    alt="Galaxy brain meme: four expanding brains paired with 'Know thyself', 'Nothing in excess', 'E', and 'If you want to talk to the Oracle, go to Delphi'."
    loading="lazy"
  />
  <figcaption>The three Delphic inscriptions plus, eventually, Borland's 1995 joke about database connectivity.</figcaption>
</figure>

Always include alt text describing the meme content for accessibility. Always include loading="lazy" because memegen images come from a third party — defer until the user scrolls there.

Workflow

When the user asks for a meme or when adding visual rhythm:

1. Identify the function. Is this a comparison (drake, pooh, bus), an escalation (gb, gru), a single thesis (cmm, philosoraptor), a categorical confusion (fry, spiderman), a "this is fine" framing, a reaction? The template carries half the joke; pick wrong and the line text won't save it.
2. Read the full live catalog, then run the random draw per "Draw against your defaults" above. Don't rely on the shortlist or on memory, and don't guess template IDs.
3. Compose the lines. Keep each line short — 3 to 8 words is ideal, anything over 12 words breaks layout. Memegen wraps text but doesn't shrink fonts dramatically; long text just looks crowded.
4. Build the URL. Use underscores for spaces, encode special characters per the rules above. Always include ?width=500 or ?width=600 for blog use — native template sizes are often huge.
5. Test the URL (recommended for production posts) by curl-ing it and checking the result is a valid PNG above 5KB. Memegen sometimes returns a tiny placeholder for invalid templates.
6. Embed with figure + alt + lazy. Don't skip alt text; meme accessibility matters.
7. Check distance from other memes. Default minimum spacing between two image memes is ~800 words. Below 600 words feels like crowding. Exception: if there are other visual elements between them (text meme, SVG, mermaid, footnote), the spacing can drop to ~500 words because the eye has rested between. The rule is "no two image memes within 800 words of each other unless the prose between them already has a different visual rest."

Density guidelines (non-serious register, default)

For a Franklin-style post (Borges-light, irreverent, philosophical):

• Posts under 800 words: 1 image meme is plenty, often zero.
• Posts 800–1500 words: 1–2 image memes, well-spaced.
• Posts 1500–2500 words: 2–3 image memes, plus other visual rests (text memes, SVG, mermaid, footnotes). Aim for the screen heuristic — visual rest every 400–500 words.
• Posts over 2500 words: 3–4 image memes maximum. Beyond that, the visual register starts dominating the prose and the post feels like a content-marketing listicle.

For serious register (when explicitly marked as such), apply the old single-image-meme rule and require full reflexive landing.

When the user wants a meme but no template fits

Three options, in order of preference:

1. Use the closest template anyway — the cultural shorthand still helps the joke land even if not perfectly fitted.
2. Custom background — memegen supports ?background=URL to overlay text on any image you provide. Useful when the writer has a specific image in mind and wants meme-style impact text on it.
3. Fall back to inline SVG meme — for galaxy-brain-shaped or drake-shaped jokes where the cultural template isn't quite right, an SVG illustration can carry the same structural humor without invoking the specific cultural reference.

API limits

memegen.link has no published rate limit and is free. For blog use this is irrelevant. The service has been online and stable since 2014. Embedding URLs in posts is a soft dependency on a third-party service — if memegen.link goes down, your memes break. Acceptable risk for a personal blog; not acceptable for high-stakes production. For mission-critical use, generate the meme once and self-host the resulting image.

What to deliver

When the user asks for a meme:

1. The complete embeddable markdown ready to paste into a post.
2. The bare URL as a separate line, in case they want to test it first or use it elsewhere.
3. One sentence explaining the choice of template — why this format carries the joke better than alternatives. Brief, not a lecture.

If proposing multiple meme options for a post, present them as numbered candidates with image + caption + reasoning, and let the user pick.

When adding rhythm-memes (not single requests, but distributing visual rests across a long post), present them as a numbered list of proposed insertion points with the meme URL and one-line reasoning for each. Let the user accept/reject individually.