The funniest AI prompts are short, specific and built around one joke: put a believable subject in an impossible situation, give it a deadpan expression, and name a visual style that suits the gag. Below are 44 copy-paste funny AI image prompts in seven groups, plus four short video prompts, a formula for writing your own, and which Kenerate model fits each type of joke (as of 24 October 2026).
What makes an AI image prompt funny?
Most AI humour comes from contrast. A cat in a business suit isn't funny because of the suit. It's funny because the cat looks completely serious about the quarterly report. When you write a funny prompt, you're looking for one of these five patterns:
| Joke pattern | How it works | Example setup |
|---|---|---|
| Role swap | An animal or object does a human job | A goose working as a traffic warden |
| Scale gag | Something is far too big or far too small | A hamster with a full-size grand piano |
| Mock-epic | A trivial moment shot like a blockbuster | A dropped sandwich as a war-movie scene |
| Wrong place | A character lands in a setting it doesn't belong in | A medieval knight in a self-checkout queue |
| Deadpan prop | A serious person holds one ridiculous object | A news anchor presenting with a baguette for a pointer |
Every prompt in this list uses one of those patterns. Pick one, not three: guides from media.io and Lilach Bullock both say that stacking too many jokes into one prompt dilutes the result. One clear twist reads instantly. Five twists read as noise.
What is the formula for a funny AI prompt?
Use this five-part structure. It works on every modern image model:
[Style] of [subject] [doing an action] in [setting], [expression], [lighting/framing].
- Style first. "Photorealistic", "1980s sitcom still", "Renaissance oil painting" or "Pixar-style 3D render". Deadpan jokes land harder in photorealism; slapstick suits cartoons.
- Subject with one detail: "a very round orange cat wearing reading glasses".
- Action verb: presenting, arguing, filing, refereeing. A verb turns a portrait into a story.
- Setting that clashes with the subject.
- Expression and framing: "completely serious expression, eye-level shot, soft office lighting".
Google's own Gemini image guide recommends the same approach for stylized images: name the style, the subject and its accessories, the activity, then the visual qualities (Google AI for Developers). Describe the scene in full sentences rather than listing keywords. If your first try falls flat, change one thing (usually the expression or the setting) rather than rewriting everything. Our guide to writing AI image prompts goes deeper on that method.
Funny animal prompts: which ones work best?
Animals in human jobs are the most reliable category because the model already knows both halves of the joke.
- Photorealistic office photo of a fluffy orange cat in a tiny navy suit presenting a bar chart titled "NAP TIME KPIs" to a boardroom of unimpressed pigeons, serious expression, fluorescent lighting.
- A golden retriever as a high-street barista, wearing a green apron, proudly handing over a coffee with a tennis ball floating in it, café interior, warm morning light.
- Wildlife-documentary photo of a raccoon in a hi-vis vest directing traffic at a busy intersection, holding a stop sign, dead serious face, overcast daylight.
- A goat on a TV cooking show, standing at a kitchen counter, eating the recipe card while the studio audience gasps, bright studio lighting, wide shot.
- A pug in a judge's robe and white wig banging a squeaky toy gavel, courtroom, dramatic low angle, film-still look.
- A penguin in a tiny tuxedo working as a hotel concierge, handing a guest a room key, luxury lobby, symmetrical framing.
- Three ducks in matching tracksuits doing a synchronized aerobics class in a 1980s gym, VHS photo look, neon leg warmers.
- A very small hamster lifting an enormous dumbbell made of two blueberries, gym mirror behind, sweatband on its head, motivational poster lighting.
- An owl as a strict librarian peering over half-moon glasses at a noisy squirrel, old wooden library, dusty sunbeams.
- A llama at a job interview, wearing a tie, sitting across from a nervous HR manager, CV on the desk, neutral office lighting.
How do I make my pet look funny with AI?
Upload a clear, well-lit photo of your pet as a reference image, then write only the change you want and what must stay the same. Say "keep the dog's face, fur colour and markings exactly the same" so the model doesn't swap your pet for a generic one.
- Turn my dog into a Renaissance nobleman in an oil portrait, velvet robe, ruffled collar, holding a chewed slipper like a royal sceptre. Keep its face and markings exactly the same.
- Put my cat on the cover of a 1990s business magazine, headline "THE DISRUPTOR", arms crossed, power pose. Keep its fur pattern unchanged.
- Make my rabbit an action-movie hero walking away from a cartoon carrot explosion, slow-motion poster style, dramatic orange backlight.
- Show my dog as a weather presenter pointing at a map where every region says "WALKIES", TV news graphics.
- Turn my cat into a grumpy medieval king on a throne made of cardboard boxes, candlelit hall.
- My parrot as a stand-up comedian at a brick-wall comedy club, holding a tiny microphone, single spotlight.
For pet edits, a multi-reference model such as Nano Banana 2.1 (up to 14 reference images) keeps the animal recognisable. On Kenerate it costs 21 credits per image at 1K.
What are some funny office and work prompts?
Work humour lands because almost everyone recognises the situation.
- A corporate stock photo of five office workers high-fiving over a spreadsheet that just says "IT SAVED", overly bright smiles, glossy stock-photo lighting.
- A medieval knight in full armour waiting on hold at an office desk, elevator music notes floating around his helmet, 1990s cubicle.
- A wizard working at an IT help desk, casting a spell on a printer that is on fire, tired expression, fluorescent lighting.
- An astronaut in a full spacesuit stuck in a revolving door at a bank, people walking around him, street photography.
- A motivational office poster of a sloth crossing a finish line in last place, caption "PROGRESS IS PROGRESS", classic poster layout.
- A team meeting where every chair is occupied by a different house plant and one human is presenting enthusiastically, documentary photo.
Which funny food prompts work?
Food characters and "dramatic food" scenes are easy wins because food has strong shapes and colours.
- A croissant detective in a trench coat interrogating a nervous donut under a single hanging lamp, film noir, black and white.
- A slice of pizza on a red carpet surrounded by paparazzi flashes, wearing tiny sunglasses, premiere photo.
- A broccoli bodybuilder flexing on stage at a vegetable bodybuilding contest, judges are carrots holding score cards.
- An epic war-movie still of a single fallen chip on a kitchen floor, the rest of the chips watching in sorrow from the bowl, smoke and dramatic lighting.
- A sushi roll riding a tiny surfboard over a wave of soy sauce, high-speed action photo.
- A birthday cake looking terrified as a crowd sings around it holding knives and forks, 1970s family photo look.
How do I write absurd product ad prompts?
Parody ads are a good use for models that render text well, because the joke usually lives in the headline. Write the exact text in quotes and describe the font.
- A glossy 1980s TV-shop advert for "THE SOCK FINDER 3000", a man in a bright jumper holding a handheld radar, bold red headline text, starburst saying "AS SEEN ON TV".
- A luxury perfume ad for "Eau de Wet Dog", elegant black bottle, golden light, a damp spaniel looking smug next to it.
- A minimal tech launch poster for an "iSpoon Pro", a single ordinary spoon on a black background, headline "We reinvented eating."
- A vintage cereal box called "MONDAY FLAKES", the mascot is a tired cartoon sun yawning, retro packaging design.
- A safety warning sign in airport style that reads "CAUTION: CAT MAY SIT ON KEYBOARD", pictogram of a cat on a laptop.
- A 1950s magazine advert for a "Personal Raincloud", a smiling woman with a tiny storm cloud above her head, pastel colours.
For text-heavy jokes, GPT Image 2.5 Sunburst (7.2 credits at medium 1K on Kenerate) and Ideogram v4 (15 credits) are the strongest picks. Check spelling on every output anyway; even good text models occasionally drop a letter.
What are good meme-format prompts?
Meme prompts need empty space for a caption, so say where the text should go, or leave room and add the caption afterwards in our AI meme generator.
- A cartoon dog sitting calmly at a kitchen table while the room around it is quietly on fire, sipping coffee, flat cartoon style, empty space at the top for a caption.
- A split image: left, a lion roaring majestically on a cliff labelled "ME IN MY HEAD"; right, a kitten falling off a sofa labelled "ME IN REAL LIFE".
- A raccoon caught in a fridge at night, eyes glowing in the camera flash, holding a slice of cheese, security-camera look with a timestamp.
- A distinguished elderly tortoise in a smoking jacket reading a newspaper with the headline "STILL LOADING", leather armchair.
- A group photo of fruit at a high-school reunion, one banana arriving very late and very brown, yearbook photo style.
What are wholesome and surreal funny prompts?
Not every joke needs a punchline. Some just make people smile.
- A tiny frog driving a huge red sports car through a busy city, pedestrians staring, street photo.
- A snowman relaxing on a sun lounger at a tropical beach, sweating slightly, sunglasses on, holiday postcard look.
- A Viking warrior very carefully doing a 1,000-piece jigsaw puzzle of kittens, cosy living room, reading glasses on.
- A grand Roman statue taking a selfie with a modern smartphone in a museum, tourists in the background.
- A cloud wearing a hard hat, holding a clipboard, inspecting a rainbow for safety, soft children's-book watercolour.
Illustrated and cartoon styles are cheap to run at volume on Seedream 5.0 Pro (13.5 credits per image at 1K/1.5K), and Kenerate Flash costs 1.5 credits per image if you want to test ten variations of the same gag before picking a winner.
Which Kenerate model should I use for each type of joke?
All prices are Kenerate credits per image at the default size, taken from the app's model config on 24 October 2026.
| Joke type | Suggested model | Why | Credits per image |
|---|---|---|---|
| Fake ads, signs, headlines | GPT Image 2.5 Sunburst | Clean text, follows long instructions | 7.2 (medium, 1K) |
| Poster and typography gags | Ideogram v4 | Text-first design model | 15 |
| Your own pet or photo | Nano Banana 2.1 | Up to 14 reference images, keeps faces consistent | 21 (1K) |
| Cartoon, anime, illustration | Seedream 5.0 Pro | Strong styles, cheap at volume | 13.5 (1K/1.5K) |
| Quick drafts of a joke idea | Kenerate Flash | Fastest, cheapest test runs | 1.5 |
| Caricature-style exaggeration | Caricature tool | Built for exaggerated faces | see AI caricature maker |
You can open any of these in the image studio at kenerateai.com/app/image. If you have a joke but your prompt is thin, paste it into the free prompt enhancer and it will add style, lighting and framing for you.
How do I turn a funny image into a funny video?
A funny still becomes a funny clip when you add one action and one beat of timing. Use your image as the start frame (image-to-video) so the character stays the same, then describe only the movement.
Comedic timing tips for video prompts:
- Set up, then pay off. Describe a calm first second, then the surprise: "holds still for a beat, then…".
- Keep it short. 4 to 6 seconds is enough for one gag. Longer clips lose the punchline.
- Use a fixed camera for deadpan jokes. A static shot makes the absurd action funnier.
- Add sound if the model supports it: a spoken line in quotes or one sound effect (a squeak, a record scratch).
Four short video prompts to try:
- The cat presenter from prompt 1 clicks to the next slide, which is blank; it stares at the camera for a beat, then slowly knocks the laptop off the table. Static camera, 5 seconds.
- The goat on the cooking show slowly chews the recipe card while the host freezes mid-sentence. Static camera, 5 seconds, studio audience gasp.
- The croissant detective slams a tiny fist on the table and crumbs fly everywhere in slow motion. Film noir, 4 seconds.
- The snowman on the sun lounger sips a drink, looks at the sun, and sinks two centimetres. Fixed wide shot, 6 seconds.
Wan 3.0 is a good value choice for short gag clips: it runs 2 to 30 seconds and costs 15 credits per second at 480p, 30 at 720p and 60 at 1080p on Kenerate, with native sound you can switch off. Draft at 480p and remake the keeper at a higher resolution. For more on models that handle comedy, see our breakdown of which AI makes funny videos from prompts, or turn a meme straight into a clip with the AI meme video generator.
How do I make my own funny AI image step by step?
- Pick one joke pattern from the table above: role swap, scale, mock-epic, wrong place or deadpan prop.
- Write the prompt with the five-part formula: style, subject, action, setting, expression.
- Choose the model by joke type: a text model for fake ads, a reference model for your pet, a cheap model for drafts.
- Generate two or three versions, then refine one detail at a time.
- Add a caption if it's a meme, leaving space at the top or bottom.
- Animate it (optional) with image-to-video and a single action beat.
What should I avoid in funny AI prompts?
- Real people. Don't make jokes about identifiable people (friends, colleagues, strangers, public figures) without their agreement. As Lilach Bullock puts it, the person posting is responsible, not the tool.
- Too many jokes in one prompt. One twist per image.
- Vague styles. "Funny style" means nothing to a model; "1980s sitcom still" means a lot.
- Trusting text blindly. Check spelling on signs and headlines before you post.
- Unclear AI edits. If an edited photo could be mistaken for a real one (especially in ads), add a small "AI-generated" note.
Sources
- media.io: Funny AI image prompt ideas
- Lilach Bullock: Funny ChatGPT image prompts
- Google AI for Developers: Gemini image generation and prompting guide
- Kenerate model credits: Kenerate app model config, checked 24 October 2026.






