# Humor and image-text grammar

A good concept starts with a small observable tension. The humor mechanism, voice, visual form and image-text relationship are separate choices. Pick each independently.

## Mechanics

| Mechanism | Trigger | Typical payoff | Trap |
| --- | --- | --- | --- |
| Incongruous reaction | Small event, enormous expression | Familiar overreaction | Famous face chosen before the situation |
| Understatement | Disaster spoken about calmly | Audience fills the gap | Corporate "we remain committed" jargon |
| Exaggeration | Repeating real friction to absurd scale | Recognizable emotional truth | Fake statistics presented as evidence |
| Reversal | Routine event gets an unlikely outcome | Surprise through expectation | Forced binary "not X, Y" wording |
| Literalization | Mental or technical pain appears as a physical scene | Immediate visual recognition | Explaining the metaphor afterward |
| Role collision | People with incompatible everyday vocabularies meet | Two cultures clash in one frame | Generic "marketing vs design" without a concrete trigger |
| False confidence | A character sounds certain while the scene undermines them | Gentle self-own or institutional irony | Misrepresenting a real person |
| Escalation | Small request grows panel by panel | Accumulated absurdity | Panels without new information |
| Affectionate recognition | A surprisingly dependable friend or sibling | Delight at ordinary behavior | Assuming all memes require suffering |
| Visual misdirection | Caption suggests one image, composition reveals another | Fast surprise | Ambiguity that needs a long explanation |

## Voices

Deadpan, weary, flustered, self-aware, smug, excited and mock-formal are delivery choices. An "angry designer" voice can use more than one comedic mechanism. Keep the chosen voice plausible for the person depicted, including dialect and seniority.

## Image-text relationships

- Reaction: caption supplies cause; a face supplies the emotion. On-image text optional
- Dialogue: caption names a speaker or trigger; in-image text is another actual speaker
- Labeling: short labels recast objects or participants; captions often unnecessary
- Understatement: picture shows high stakes; short text plays them down
- Literalization: visual embodies an abstract feeling or technical operation
- Delayed reveal: top panel sets one expectation; lower panel changes its meaning
- Intentional echo: the same phrase repeats because the second speaker or new context changes its meaning
- Text-free: a photo, screenshot, gesture or sequence can carry the full joke; optional accessibility alt text is external to the image

The two-mouth test applies to dialogue. For labeling, repeated words can be the joke. Text-free concepts do not need a fictional line inside the frame. The relation should add a second piece of meaning unless repetition is intentionally doing the work.

## Format selection

Reaction still, original actor scene, safe original illustration, staged screenshot, fake chat clearly marked as a constructed scene, accurately drawn UI, object labels, split panel, short video or pure-text joke are valid. A named template is a search cue unless usage rights are documented. Provide an original fallback that keeps the same mechanism without copying a recognizable character.

## Batch planning

For three concepts, prefer three different moments and at least two different mechanisms. Do not force numeric receipts, pain, a colon, a pun or a famous face into every concept. If the user explicitly requests variations of one concept, keep the same caption across alternate visual options so the comparison is meaningful.

## Quick counterexamples

Weak: "لما الشغل يبقى صعب" with an unrelated crying celebrity

Stronger illustrative setup: "لما الـ QA يرجعلك نفس الـ bug بعد ما قفلت التذكرة الساعة 11:57" with an original weary character staring at the green "resolved" label

Different direction: a staged chat of two colleagues who both think the other owns the same approval, with no reaction face

These are fictional examples of form, not claims about actual workplace events.