What You're Trying to Make
One selfie in, and out comes an adorable, pocket-sized version of you. Your tiny chibi self stands on a giant open palm — real face, real hair, shrunk to a big-headed little figurine — breathing, blinking, alive at miniature scale. Then a giant fingertip reaches in and boops your cheek, and your tiny self reacts: it flinches, puffs its cheeks, swats the finger away with both little arms, and pouts... before the finger gently pats its head and it softens into a hug.
That's the whole video — hyper-realistic macro photography of a miniature you, creamy bokeh behind, a real contact shadow under your tiny feet. The appeal is pure scale-and-cute: a living pocket person being played with like a toy. This guide covers the whole thing — where the trend came from, the exact prompt recipe (the shrink-down, the palm, the finger boop, the reaction), the one artifact that ruins most attempts, and the one-tap way to make yours.
Fastest way — Tiny Me in My Hand on Starrd bakes the whole thing into one tap: upload a selfie and it renders the shrunk-down you, the palm, the cheek boop, and the grumpy-then-hug reaction. First video free, no prompt to write. Want the full method first? Read on. ↓
Is This a Trend? (Yes — the "Mini Me in Your Palm" Format)
It's real and it's everywhere. On TikTok and Instagram it goes by a few names — "mini me," "tiny person in hand," "miniature me," "pinch cheek" — but it's always the same beat: a normal photo becomes a tiny, hyper-realistic figurine of that person, resting in someone's palm, and a giant hand plays with them while the little figure reacts.
Why it works: it hits the exact same nerve as tiny food, dollhouses, and pocket pets — the irresistible cuteness of a fully-detailed thing shrunk to miniature. Add a face that's recognizably you and a reaction (grumpy swat, shy giggle, a hug) and it stops being a static figurine and becomes a character with a personality. It's endlessly personal because the mini me is you, and it's endlessly remixable because you control the reaction and what the giant finger does. Most versions are made in a couple of taps with a viral effect — but the good ones live or die on two things: keeping your real face on the miniature, and getting the finger to touch the cheek without smearing it.
The Fastest Way — Use the Tiny Me in My Hand Template on Starrd
The template is live in the Starrd library. It packages every step in this guide — the shrink-to-chibi, the giant palm, the cheek boop, and the grumpy-then-hug reaction — into a single upload.
- Pick one clear selfie. Front or three-quarter view, eyes open, good lighting.
- Open Tiny Me in My Hand in the Starrd app or web library.
- Tap generate. It shrinks you into a palm-sized chibi, sets you on the giant hand, boops your cheek, and renders a ~10-second 9:16 vertical clip on Seedance 2.0 — tiny squeaks and room tone included.
One selfie, a few minutes. No prompt writing, no model picking. Want a different vibe? Set it in Director's Notes:
- The reaction — grumpy-and-swatting by default; you can make it giggly and delighted, startled-then-posing, or full sass.
- The outfit — keep your photo's look or restyle the mini me.
- The prop — the giant hand can boop your cheek, pat your head, or offer a tiny high-five.
Tiny Me in My Hand
Upload one selfie and get shrunk into an adorable palm-sized version of yourself — your tiny chibi self stands on a giant hand, a fingertip boops your cheek, and you swat it away and pout before caving into a hug. Set the reaction and outfit in Director's Notes. No prompt to write. First video free.
The rest of this guide is for people who want to roll their own — control the reaction, the styling, or run it on a different model.
Or, Build It Yourself — What You Need
Three things:
- One clear selfie of yourself. Front or three-quarter view, eyes open, good lighting.
- An AI video model that takes a reference image and generates its own audio. Seedance 2.0 is the safest; Kling and Veo also work.
- A platform to post on. TikTok, Reels, and Shorts are where this travels.
You don't need any real footage, a green screen, or a single prop.
Step 1 — Pick Your Reference Photo
The selfie you feed the model is the face that ends up on the miniature. Because the whole payoff is "that tiny figurine is actually me," this choice matters more here than in almost any other trend.
Use:
- A clear, well-lit photo of just you
- Front-facing or three-quarter angle, eyes open
- A head-and-shoulders or upper-body shot (the model keeps your face and hair, shrinks the rest)
- Visible, distinct hair — it's a big part of what makes the chibi read as you
Avoid:
- Group photos (the model doesn't know who to shrink)
- Sunglasses, hats, or anything covering the face
- Low-resolution or motion-blurred shots
- Heavy beauty filters (they wash out the very features that make the mini me recognizable)
Step 2 — Lock the Concept
Decide the details before you write the prompt:
- 🤏 The scale: palm-sized. Your tiny self should stand on an open hand with a real contact shadow and correct weight — that grounding is what makes it look like a photograph, not a paste-in.
- 🧍 The look: an adorable chibi figurine with exaggerated proportions (a big head, roughly 1:3 head-to-body) but your real face, skin, and hair. Hyper-realistic, not cartoon.
- 📷 The camera: macro photography — the hand and figure razor-sharp, the background melted into creamy bokeh, soft warm indoor light. This is the "expensive miniature photo" look.
- 😾 The reaction: the personality beat. Grumpy-swat is the classic (and funniest); giggly, shy, or sassy also work. This is where the tiny figure becomes a character.
Step 3 — Write the Prompt
A mini-me video lives on scale + a real reaction, shot like macro photography. Build a clean anchor frame first — a single still of the tiny you on the palm — then let the video move through the beat. Copy this and swap in your specifics:
ANCHOR FRAME (single still): a hyper-realistic macro photograph, vertical 9:16. A palm-sized miniature of the person from the uploaded photo — keeping their exact real face, skin, and hair, shrunk into an adorable chibi figurine with big-head 1:3 proportions — stands upright on a giant open, upturned human palm. From the side of frame, a second hand's giant index finger reaches toward the tiny figure. Realistic contact shadow under the tiny feet, correct miniature scale. Macro lens: the hand and figure razor-sharp, the background a creamy blurred bokeh, soft warm indoor light. Photorealistic, full-frame, no text.VIDEO (10s, vertical 9:16, hyper-realistic macro photography, creamy bokeh):[0–2s] The tiny figure idles on the palm — a slow blink, subtle breathing, a little stretch — as the camera pushes in.[2–4s] A giant fingertip lightly BOOPS its cheek and immediately lifts fully clear; the cheek dimples slightly and springs right back. The tiny figure startles and pouts.[4–6s] It swats the finger away with both little hands, cheeks puffed, grumpy.[6–8s] The fingertip gently returns and pats the top of its head; the tiny figure flinches, then hesitates.[8–10s] It softens and wraps both little arms around the fingertip in a warm hug, beaming a happy smile. Gentle pull back.Audio: soft indoor room tone, tiny high-pitched squeaks and a little annoyed grumble, a light playful "boop" on contact, a soft "aww." Keep the top of the mix open for a trending sound.Keep the face clearly the person from the photo. The finger must never stick to or drag the cheek — it's a quick light touch that lifts clean.
The move that makes or breaks this one is the cheek boop, not a pinch. AI models have no real physics, so if you tell them to "pinch" the cheek, the finger grabs and drags the skin into taffy as it pulls away. Prompt a light quick boop that immediately lifts fully clear and the cheek springs back, and add negatives — "cheek sticking to finger, skin stretching, melting." That one change is the difference between adorable and cursed.
The non-negotiable elements:
- Correct miniature scale — palm-sized, with a contact shadow. Grounding sells the illusion.
- Your real face on a chibi body — big head, real features. Not a cartoon, not a generic doll.
- Macro bokeh — sharp hand and figure, melted background. This is the whole "photograph" look.
- A clean boop + a real reaction — the finger lifts clear, the tiny you reacts with personality.
Step 4 — Pick a Model
- Seedance 2.0 — what Starrd runs on. Holds your face through the shrink-down, keeps the scale believable, animates the boop and reaction as one continuous macro shot, and generates the squeaks and room tone in the same pass. Safest pick.
- Kling — strong motion; push the scale and "real human face, not a doll" hard or the miniature drifts generic.
- Veo 3.1 — capable with synced audio; keep "macro lens, creamy bokeh, palm-sized" explicit or it flattens the depth.
- Runway Gen-4 — solid look, but the timed boop-and-react beat is harder to land cleanly in one pass.
No preference? Start with Seedance 2.0. (See our Seedance 2.0 Prompt Guide.)
Step 5 — Generate and Iterate (Common Failure Modes)
First generations rarely nail it. The usual failures and fixes:
The cheek stretches, melts, or sticks to the finger. The signature failure. Swap "pinch" for a "light quick boop that immediately lifts fully clear, cheek springs back," and add "cheek sticking to finger, skin stretching, melting, taffy" to your negatives.
The mini me doesn't look like you. Two causes: a weak reference photo, or a prompt that over-describes the face. Use a sharp, front-facing selfie and let the photo carry the identity — don't describe eye shape or skin tone in words; those fight the photo and genericize it.
The scale looks wrong (giant person, or a flat paste-in). Force it: "palm-sized miniature, correct scale, realistic contact shadow under the feet, macro lens." No shadow and no depth is what makes it look pasted on.
It looks like a cartoon. You leaned too far into "chibi." Keep the proportions exaggerated but the render hyper-realistic: "real skin texture, real hair, photorealistic macro photograph."
Nothing happens / it's boring. One poke with a long pause is dead air. Give it a beat: idle → boop → swat/pout → a second touch → a payoff (a hug, a sassy pose). Keep the pacing tight.
Budget 3–5 generations before a keeper.
Step 6 — Post It
Let the cuteness carry it — and add a trending sound. The prompt keeps the top of the mix open so you can drop a trending audio over the tiny squeaks. A soft or funny sound amplifies the "aww."
Caption it as the character. "he did NOT like that" or "pocket-sized me has an attitude" travels further than "I made this with AI." Play it as if the tiny you is a real little diva. (Within each platform's required AI-disclosure labels.)
Vertical, and let the boop hit early. It's a 9:16 format — the cheek boop around second 2–3 and the reaction are the beats a scroller stays for, so don't bury them behind a long idle.
Common Mistakes That Tank Your Video
- A pinch instead of a boop. The dragged, melting cheek is the #1 way this goes from cute to cursed. Light touch, finger lifts clean, cheek springs back.
- A weak reference photo. Blurry, dark, filtered, or group shots = a mini me that doesn't look like you, which kills the whole point.
- Wrong scale. No contact shadow, no depth, or a figure that's too big reads as a flat paste-in. Palm-sized with a shadow.
- Cartoon render. Exaggerated proportions are good; a cartoon finish isn't. Keep it hyper-realistic.
- Dead air. A single poke with nothing around it is boring. Give the tiny you an arc and a payoff.
Frequently Asked Questions
What is the mini me AI trend? One selfie becomes an adorable palm-sized figurine of you, standing on a giant hand. A fingertip boops your cheek and your tiny self reacts — usually grumpy, then a hug.
How do I make one? Prompt a model with the shrink-to-chibi / palm / cheek-boop / reaction recipe plus your photo — or use the Tiny Me in My Hand template and just upload one selfie.
Why does the cheek stretch weirdly? Because the model has no physics and drags the cheek as the finger pulls away. Use a quick "boop" that lifts clean, not a "pinch."
How long should it be? About 8–12 seconds. Short loops better and holds attention; long clips run out of beat.
Do I need to label it as AI? Yes — all major platforms require it, and a living pocket-sized you is obviously staged anyway.
Related Reading
- How to Make a Netflix Documentary AI Video — the deadpan self-insert: you as the dead-serious subject of a fake documentary.
- How to Make a Red Carpet AI Video — the paparazzi flex, another one-selfie self-insert.
- How to Make an AI Squish Video — the satisfying-squish cousin, if you liked the cheek-boop physics.
- Viral AI Video Trends (2026): The Monthly Roundup — every trend worth making this month, each with a one-tap template.
- Seedance 2.0 Prompt Guide — the framework behind the timed-beat prompt above.