DesignNovember 12, 2026, 10:31 AM·11 min read

Funny Gemini Prompts: 12 Ready Prompts for Your Own Photos (2026)

Funny Gemini prompts: a tiny you on the rim of a tea glass, a cat running a meeting, a giant cat next to the Maiden's Tower, a wildlife documentary still, a court portrait, meme text and a video of a cat knocking a glass over. Every line is backed by numbers from our earlier Gemini tests.

Funny Gemini Prompts: 12 Ready Prompts for Your Own Photos (2026)
Key Takeaways
  • The 12 funny Gemini prompts in this post rest on four patterns: playing with size, putting an ordinary moment into a serious scene, exaggerating one thing, and meme text. All of them work with your own photo or a photo of your cat.
  • With a short "make it funny" prompt, the model fills the gaps on its own: in our 3D figurine test, 8 of 20 images got made-up text; with a "no text, no logos" sentence on the first line, 1 of 12.
  • Name the hair, beard, glasses and moles on your face one by one. A generic "keep moles and glasses the same" list makes the model add features you don't have: in our emoji test, 8 of 11 images got a mole that wasn't there.
  • Put the meme text in quotes and say "copy exactly". In our invitation test, 65 cards written this way had no wrong letters.
  • Funny videos in Gemini are made with Omni and need a paid Google AI plan. According to Google's documentation, speech is fully supported only in English; for a single scene you have to write "one continuous shot".

The 12 prompts in this post rest on four patterns: making a person or a cat absurdly big or small, putting an ordinary moment into a very serious scene, exaggerating one thing, and stamping meme text on the photo. Each time you upload the photo to Gemini, write the scene and the thing to exaggerate, and ask for everything else to stay the same.

The list has a tiny you on the rim of a tea glass, a cat running a meeting, a giant cat next to the Maiden's Tower, a wildlife documentary still, a court portrait holding a tea glass, exaggerated muscles, you at 80, meme text and a video of a cat knocking a glass over. Under each prompt we explain why each line is there, with numbers from our earlier Gemini tests.

Source: our earlier tests, October 15 - November 11, 2026

We did not test these 12 prompts as a separate set: the rest of our November test budget went to cover images, and we have no API key to run ChatGPT's image model. The lines in the prompts come from our earlier tests in the Gemini API with Nano Banana 2 Lite, Nano Banana 2 and Nano Banana Pro, all run with Turkish prompts: face preservation (November 3, 122 images), 3D figurine (October 29, 133 images), invitation text (October 28), snow trend (November 4, 81 images), emoji (November 5, 123 images) and men's portrait (November 11). The people in the tests were fictional, AI-generated people. The tiny man on the cover is fictional too; we had Gemini draw him from a description.

What Makes an Image Funny? Four Patterns

In each of the four patterns, one part of the image doesn't fit the rest: the size, the setting, a body part, or the text on top. That is why the prompts put the exaggerated thing on a single line and the things that must stay the same on separate lines.

Table of the four patterns for funny Gemini photos: playing with size, serious scene, exaggeration and meme text, with the line each prompt has to include
The four patterns and the line every prompt includes for each one
PatternExampleLine the prompt always includesPrompt
Playing with sizeTiny you, giant catSize measured against an object: "as tall as the glass", "twice as tall as the tower"1, 4
Serious scene, ordinary momentCat running a meeting, documentary, court portraitThe scene's technique: telephoto lens, oil paint, office lighting3, 5, 6
Exaggerating one thingMuscles, afro, age 80That everything outside the exaggeration stays the same7, 8, 10
Meme textTop and bottom textThe line in quotes, plus "copy exactly"9
Funny videoCat knocking a glass over"One continuous shot" and a second-by-second flow11

In the size prompts we gave height against an object ("as tall as the glass", "twice as tall as the tower"). We did not measure this size description on its own; it rests on the October 16 pose test, where a physical description worked better than naming the pose.

How to Make a Funny Photo in Gemini: 4 Steps

  1. Open gemini.google.com or the Gemini app and sign in with your Google account. You have to be at least 18 to edit a photo you upload.
  2. Add a waist-up photo with your face clearly visible using the + icon. For your cat or dog, pick a shot where it is alone and its eyes are open.
  3. Copy a prompt from below, fill in the square brackets and send it in the same message as the photo.
  4. If the result isn't what you wanted, type Prompt 12 in the same chat and list only what came out wrong. Save the image you like with the download icon on it.
Don't upload a full-body photo

In the November 3 test, when the face shrank to 10-23% of the frame, the similarity score fell from 0.76 to 0.66 and moles were erased in 4 of 6 images. Start with a waist-up photo and let the prompt build the full-body scene.

12 Ready-to-Use Funny Gemini Prompts (Copy and Paste)

Fill in the square brackets to match your own photo. If something isn't on your face (glasses, beard, a mole), delete that part entirely. The original prompts are Turkish; this English version can also be pasted into ChatGPT as is. We did not test the ChatGPT side.

1) Tiny you on the rim of a tea glass

Prompt 1 : Tiny me
Prompt
No text, logos or labels anywhere in the image. Place the person from the photo I uploaded into a realistic photo as a tiny person the size of a palm.
Scene: a light-colored wooden kitchen table, morning light. On the table is a handleless, slim-waisted Turkish tea glass on its saucer; the person sits on the edge of the saucer and is as tall as the glass.
The person hugs a single sugar cube with both arms, straining as if carrying a boulder; cheeks puffed out, eyebrows raised.
Face as in the photo: [hair color and length], [beard or mustache shape, if any], [glasses frame, if any], skin tone [color]. Do not add moles, glasses or a beard that are not in the photo.
Clothing: [top and its color]. The watch and bag from the photo are not in this shot.
A normal-size teaspoon stands next to the person and looks bigger than them. Landscape 16:9.

We gave the size with an object: "as tall as the glass". In our October 16 pose test, a physical description worked better than naming the pose (3 fully correct out of 6 tries, against 1). The watch line comes from November 3: when a person moved into a new scene, the wristwatch came along in 19 of 32 images; writing "not in this shot" brought that to 0 of 6.

2) Short version: a funny photo from one idea

Prompt 2 : Short funny photo
Prompt
Do not add text, logos or watermarks to the image. Make this photo funny: [funny idea, e.g. show me riding on my cat's back like a horse].
Keep my face, hair and clothes as in the photo; do not add glasses, moles or accessories that are not in the photo.
Make it look like a real photo.

Bring the idea yourself instead of typing "make it funny" and leaving it to the model. In the October 29 3D figurine test, the short prompt put a made-up name or text on the box and base in 8 of 20 images; the line "no text, names or logos" cut that to 1 of 12. That line comes first here because a ban we put on the last line on October 15 did not hold.

3) A cat running a meeting

Prompt 3 : Boss cat
Prompt
No text, logos, name cards or screen text in the image. Put the cat from the photo I uploaded at the head of a realistic office meeting.
The cat sits upright in a chair at the head of a long meeting table; its front paws are on the table, with an empty notepad and a glass of tea in front of it.
Dress the cat in a navy suit jacket and a white shirt; no tie.
Fur as in the photo: [color], markings [position from the viewer's side, e.g. black patch above its left eye]. Whiskers and eye color [color] unchanged.
Four chairs stand along the two sides of the table, all four empty. The cat's face is very serious, as if frowning.
Landscape 16:9, real office lighting.

Describe where the markings are "from the viewer's side": on October 29 this description put a dog's markings in the right place in 6 of 6 images. We spelled out the tie because in the November 11 men's portrait test the word "suit" brought a tie along on its own in 2 of 2 tries.

4) A giant cat next to the Maiden's Tower

Prompt 4 : Giant cat
Prompt
No signs, posters, text or logos in the image. Place the cat from the photo I uploaded into a realistic photo as a giant bigger than the city.
Location: the Bosphorus in Istanbul, in front of the Maiden's Tower. The Maiden's Tower is a white stone building with a single tower, standing on a small rocky islet in the sea. The shot is taken with a wide angle from the stone quay on the shore.
The cat sits in the sea and is twice as tall as the tower; it reaches one paw curiously toward the top of the tower.
Fur and markings as in the photo: [color], [position of the markings from the viewer's side].
Three small people stand on the quay looking at the cat, backs turned, faces not visible.
Landscape 16:9, late afternoon light.

Next to the place name we wrote what the building looks like. In the November 10 plan comparison, Nano Banana Pro turned a place name given without a description into a different building; in the November 4 snow test, Galata Tower was recognized in 8 of 8 images. If you swap in a landmark from your own city, do the same: the name, then its color and shape.

5) A wildlife documentary still

Prompt 5 : Documentary still
Prompt
No text, channel logo or caption bar in the image. Show the person from the photo I uploaded as a still from a wildlife documentary.
Keep the person in the same spot and pose as in the photo: [e.g. asleep on their side on the couch].
The shot looks secretly taken from far away with a telephoto lens: out-of-focus leaves and branches in the foreground, the person in focus, slight grain.
The light is dim and warm, like early evening. Keep the objects in the room where they are; do not add new people or animals.
The person's face, hair and clothes as in the photo; do not add glasses or moles that are not in the photo.
Aspect ratio same as the photo.

The joke is in how serious the shot looks: the more convincing the telephoto lens, the leaves and the grain, the funnier it gets. The channel logo line comes from an October 29 finding: the model drew logos on objects that carry logos in real life (laptop lids, monitors). TV documentaries always have a logo in the corner too.

6) Court portrait holding a tea glass

Prompt 6 : Court portrait
Prompt
No signature, text, frame label or museum sign in the image. Turn the person from the photo I uploaded into a court portrait in oil from the 1500s.
The canvas reaches all four edges of the image; the background is a dark green velvet curtain, the person is shown from the waist up and looks at the viewer from a three-quarter angle.
The person wears dark red velvet period clothing with a white ruffled collar.
Instead of a gold goblet, they hold [funny object, e.g. a slim-waisted Turkish tea glass]; the face is very serious.
Brushstrokes are visible, the light comes from the left, the colors are dark and warm.
Face as in the photo: [hair color and length], [beard or glasses, if any]. Do not add moles or glasses that are not in the photo.
Portrait 4:5.

We described the materials and the light and named no painter. In the October 22 drawing test, a short style prompt produced an unexpectedly colorful picture, while the prompt describing the materials and the paper held up. The "all four edges" sentence cuts the frame off in a flat print (polaroid, October 29); if you want a painting hanging on a wall, delete that sentence.

7) Exaggerated muscles at the gym

Prompt 7 : Muscular me
Prompt
No text, logos or brands in the image; the clothes are plain with no print. Put the person from the photo I uploaded into a realistic gym photo with an exaggerated, cartoon-hero muscular body.
The person stands in the middle of the gym facing the viewer, flexing their biceps with both arms; the arms and chest are bigger than any real human's could be.
Face as in the photo and normal-sized for the body: [hair color and length], [beard or mustache], [glasses], skin tone [color]. Do not add moles or tattoos that are not in the photo.
The facial expression is very self-assured: mouth closed, a slight smile, no teeth showing.
They wear a plain gray sleeveless T-shirt. Portrait 4:5.

Only the muscles are exaggerated; the face stays "normal-sized". The expression line comes from November 3: with a short prompt, Nano Banana 2 turned a closed mouth into a toothy smile in 5 of 6 images, and with "mouth closed, no teeth showing" in the prompt, 0 of 9. Gym machines and sportswear carry brands in real life, which is why the first line is there.

8) You at 80

Prompt 8 : Me at 80
Prompt
No text or date stamp in the image. Show the person from the photo I uploaded at age 80; keep the framing, pose and background as in the photo.
Aging: instead of [current hair color], thin, snow-white hair; deep wrinkles on the face, bags under the eyes.
The facial features belong to the same person: eye color, nose shape and [moles, glasses, if any] are kept.
Same clothes, with a beige cardigan added on top; in their hand is [funny object, e.g. a game controller].
The facial expression is grumpy: eyebrows furrowed, mouth closed.

The object in your hand makes the joke: the face of a grandparent scolding the grandkids, with a game controller in hand. The date stamp line comes from the October 30 test of 80s-style photos: a prompt saying "like it came out of an old album" added a stamp and a border on its own.

9) Meme text (top and bottom caption)

Prompt 9 : Meme text
Prompt
Add meme text to the top and bottom of the photo I uploaded; do not change anything in the photo itself.
Top text, in capital letters, copy exactly: "[TOP TEXT]"
Bottom text, in capital letters, copy exactly: "[BOTTOM TEXT]"
The font is bold, the letters are white with a thin black outline, centered.
Apart from these two lines there is no text, logo or watermark in the image.

Give the line in quotes and say "copy exactly". In the October 28 invitation test, 65 cards written this way had no wrong letters. On October 23, Nano Banana 2 garbled the Turkish dotless I once and got it right after a second message. Don't add a list of letters (something like "watch out for the Turkish S with a cedilla and the soft G"): in the November 6 tattoo test, a list like that made the model add a letter that wasn't in the text.

10) Giant afro joke

Prompt 10 : Afro hair
Prompt
No text or logo in the image. Change only the hair of the person in the photo I uploaded; the face, clothes, background and framing stay the same.
New hair: a giant afro, twice as wide as the head, jet black, round and fluffy. [funny object, e.g. a pencil and a teaspoon] is stuck in the hair.
No hat or bandana; all of the hair is visible.
The facial expression is as in the photo; [beard, glasses, moles, if any] stay exactly the same.

The hat line comes from the October 30 newborn test: with "bare head" the model thinned or erased the hair in 3 of 36 images, while a sentence saying there is no hat and the hair is visible gave full hair in 6 of 6. Don't use this prompt on a photo of someone wearing a headscarf: on October 30, a trend prompt asking for "voluminous hair" removed the headscarf in 6 of 6 images. Use Prompt 1, 5 or 6 instead, and in the hair line write the headscarf's color and how it is tied: on November 4, with that line the headscarf kept its color in 3 of 3 attempts.

11) Funny cat video (Gemini Omni)

Prompt 11 : Cat knocking a glass over (video)
Prompt
One continuous shot, no scene cuts. Turn the cat in the photo I uploaded into a realistic 8-second video.
The camera is static; the spot where the cat sits and the room are as in the photo.
Seconds 0-3: the cat stares seriously at the camera.
Seconds 3-6: without taking its eyes off the camera, it slowly pushes [object, e.g. the empty tea glass] on the table toward the edge with one paw.
Seconds 6-8: the glass falls out of frame, and the cat keeps looking at the camera, completely unbothered.
Audio: only room tone and the sound of the glass falling; no speech and no music. No text or logo in the video.

In Gemini this prompt needs a paid Google AI plan. Timing makes the joke; why we left speech out is explained in the video section below, in Google's own words. We did not test the video prompt.

12) Dial the exaggeration (same chat)

Prompt 12 : Dial the exaggeration
Prompt
In this image change only this: [e.g. make the muscles half as big / make the cat's expression more surprised].
Keep everything else the same: face, hair, clothes, background and framing.
These must match the photo on the face: [hair color and length], [beard], [glasses], [mole]; if they changed, fix them.
Do not add any new text, logos or people.

Type it in the chat where the image was made. On November 3, the message "you changed my face, fix it" left the similarity score at 0.72, while a correction that listed the features one by one raised it to 0.84. If the face starts drifting after two or three rounds, start over in a new chat with the original photo: on October 27, edit after edit shifted the person's identity.

If your request is refused

Check the photo first. A photo of a well-known person can get the request refused; Google's Omni documentation also says that uploading and editing images of some recognizable people is not supported. If there is a child in the photo, ChatGPT may refuse the request: according to OpenAI's GPT-4o image generation system card, a classifier estimates whether an uploaded photo contains a child. If the prompt names a real brand or channel, take the name out and describe how it looks instead.

What Went Wrong in Our Earlier Tests?

Most lines in these prompts are there to prevent two problems: text or objects the model adds on its own, and small marks on the face drifting. Each pair in the chart comes from a test where we did the same job with a short prompt and with the relevant line added.

Rate of unwanted additions in earlier Gemini tests with a short prompt and with the relevant line added, three findings side by side
Short prompt vs. prompt with the relevant line added: rate of unwanted additions (Gemini API, October 29 - November 5, 2026)
  • Made-up text (3D figurine, October 29): 8 of 20 images with the short prompt, 1 of 12 with the line "no text, names or logos in the image".
  • A mole that isn't there (emoji, November 5): 8 of 11 images with a generic "keep moles and glasses the same" list, 1 of 16 with a person-specific description and the line "do not add moles that are not in the photo".
  • Wristwatch carried into a new scene (face preservation, November 3): 19 of 32 images, 0 of 6 with the line "no watch in this shot, hands empty".
  • Toothy smile (Nano Banana 2, November 3): 5 of 6 images with the short prompt, 0 of 9 with the line "mouth closed, no teeth showing".
  • Readable made-up sign (snow trend, November 4): 4 of 4 when we asked for "blurred shop signs", 0 of 4 with the line "no shops, signs or text".

These numbers come from realistic edits and figurines. A funny scene has the same weak spots (signs, screens, print on clothes, small marks on the face), but we did not measure the rates for this post's prompts separately.

Funny Gemini Video Prompts: What to Expect From Omni

Omni is Gemini's model for turning a photo into a video. In the Gemini app it needs a paid Google AI plan; videos run 3 to 10 seconds, up to 15 seconds on AI Ultra (Google help page, November 11, 2026). According to Google's developer documentation, the model tries to build a video with several shots by default; for a single scene you have to write "one continuous shot" and "no scene cuts". That is why Prompt 11 opens with that sentence.

English (EN) is fully supported, but other languages have not been evaluated, so they may work but results can vary.

· Google AI Studio, Omni documentation (accessed November 12, 2026)

If you want a talking cat video, this sentence matters: Turkish speech may work, but Google says it has not evaluated it, and the same goes for every language other than English. That is why Prompt 11 has no speech and the joke rests on the moment the glass falls and on the cat's stare. If you want a spoken line in Turkish or another language, add the audio later in Instagram or CapCut.

Can You Make Funny Photos With Free Gemini?

Yes. Every prompt except the video one works on an account without a plan. According to Google's help page, images download at 1K without a plan and at 2K with a Google AI plan; paid plans also get a "Redo with Pro" option. Most requests from accounts without a plan go to Nano Banana 2 Lite. In the November 3 test, with the prompt that described the face feature by feature, the three models scored almost the same on similarity (Lite 0.765, Nano Banana 2 0.767, Pro 0.768); the gap opened with the short prompt, where Lite stayed at 0.71. On an account without a plan, pick the detailed prompts over the short version (Prompt 2).

Common Mistakes

  • Typing "make it funny" and stopping there. The model picks the joke itself and puts text on empty surfaces; bring the idea yourself.
  • Writing a place name on its own. In the November 10 test, Nano Banana Pro turned a place name given without a description into a different building; add its color and shape next to the name.
  • Asking for a "blurred" sign. The model draws a readable but meaningless sign; write "no shops, signs or text".
  • Asking for a crowd of objects by number. On October 20, "exactly 16 candles" came out with too many in 7 of 8 tries; for more than 10, describe a layout such as "4 rows of 4".
  • Using a hair prompt on a photo with a headscarf. Prompt 10 can remove the headscarf; pick Prompt 1, 5 or 6 and state the headscarf's color separately.
  • Adding a letter list for meme text, or writing only "fix it" when correcting. Give the line in quotes, and when correcting, list what came out wrong one by one.

To generate images in Gemini you have to be at least 13, and to edit a photo you upload, at least 18 (Google help page, November 11, 2026). The same page asks you not to violate other people's copyright and privacy rights.

Under KVKK, Turkey's personal data protection law, a photo of someone's face is personal data. Sharing a documentary-style shot of your dad asleep on the couch in the family group chat is one thing; posting it to a public account without his permission is another. Ask before you upload. Don't upload photos of celebrities: the request may be refused, and the Omni documentation does not support images of some recognizable people either. Keep funny shots made from your child's photo within the family.

Frequently Asked Questions

The photo prompts are. Upload the photo to Gemini with your Google account and paste the prompt; on an account without a plan, images download at 1K. The funny video (Prompt 11) is made with Gemini Omni and needs a paid Google AI plan.

Open Gemini, upload a waist-up photo where your face is clearly visible, copy one of the prompts in this post, fill in the square brackets and send it in the same message as the photo. Fix the part you don't like in the same chat with Prompt 12.

Start with "One continuous shot, no scene cuts", describe the video second by second and write the audio as a separate line. According to Google's Omni documentation, the model builds several shots by default, and speech is fully supported only in English. If you want a line in another language, add it afterward as audio.

Yes. Prompt 3 puts your cat at the head of a meeting, Prompt 4 puts it next to the Maiden's Tower, and Prompt 11 makes the video. Describe where the markings are from the viewer's side: in our October 29 test, that description put the markings in the right place in 6 of 6 images.

With a short prompt, the model redraws the face and fills in any marks you didn't mention its own way. Name your hair, beard, glasses and moles, and leave out what you don't have. If your face changed, fix it in the same chat by listing the features: in our November 3 test, that correction raised similarity from 0.72 to 0.84.

You can paste them into ChatGPT as they are. The numbers in this post come from Gemini tests run with the original Turkish prompts; we did not test ChatGPT for this post. The video prompt (Prompt 11) was written for Gemini Omni.

We drew the tiny man on the cover with an English prompt similar to Prompt 1. For the glass we wrote "slim-waisted" and said nothing about a handle; Gemini added a handle. That is where the word "handleless" in Prompt 1 comes from: any detail you don't want the model to decide for you, write into the prompt yourself.

We Offer This Service

Instagram and Facebook Ads Management

A funny post can bring followers to your account; turning followers into customers takes ads and measurement. We manage your Instagram and Facebook ads with tracking in place. Let's look at your account together in a free initial call.

Design
Write to us