How to Make a 'Give Me Peanut Butter Please' AI Video (2026 Trend)

What You're Trying to Make

One selfie in, and out comes the most unhinged clip on the internet: your real, unchanged human face on a creature whose entire body is glistening, creamy peanut butter — held up on long, spindly, daddy-long-legs limbs. It skitters down a real city sidewalk on shaky vertical phone-cam footage, snatches a peanut butter jar out of a stranger's grocery bag, pours the whole thing over its own head, spins in place smearing a peanut-butter ring on the pavement, then bursts into a grocery store and rampages the aisles — knocking jars off the shelves — before scrambling up a giant mountain of peanut butter. The whole time it screams "give me peanut butter please" in a high-pitched voice, on a loop.

It's absurd, it's a little unsettling, and it's very funny — that's the whole recipe. This guide covers the trend, the exact prompt to build it yourself, and the one-tap way to make yours.

Fastest way — upload one selfie and Starrd bakes the whole thing into one tap: it builds the peanut-butter creature with your real face and renders the full 12-second rampage, audio and all. First video free, no prompt to write. Want the full method first? Read on. ↓

Is This a Trend? (Yes — It's the Breakout of the 'AI Cast' Wave)

It's real, and it's huge. The searched phrase is exact — people type "give me peanut butter please" into TikTok looking for it — and the format crossed 5.6 million posts in 2026, tied to the streamer idksterling. It's the breakout entry in the broader "AI Cast" trend: absurd body-horror comedy clips where AI casts a person as an impossible creature — made of peanut butter, spaghetti, melting wax — that keeps their real face and does something chaotic in an ordinary real-world setting, shot like someone caught it on their phone.

Here's why this one won the group: it has a catchphrase. "Give me peanut butter please," screamed in a high-pitched voice on a loop, is instantly quotable and instantly re-createable — every viewer immediately imagines their own face on the creature. Stack that on the grocery-store rampage (a familiar, boring location made insane) and the uncanny preserved face, and you get a clip that's equal parts funny and cursed — the exact blend the For You page rewards. Trends like this move fast, so the quickest path to a clip is almost always a ready-made template.

The Fastest Way — Use the Give Me Peanut Butter Please Template on Starrd

The Give Me Peanut Butter Please template is live in the Starrd library. It packages every step in this guide — preserving your exact face, building the wet peanut-butter body with spindly legs, and animating the full street-to-grocery-store rampage with the screamed catchphrase — into a single upload.

1. Pick one clear selfie. Front or three-quarter view, eyes open, good lighting.
2. Open the Give Me Peanut Butter Please template in the Starrd app or web library.
3. Upload and tap generate. The template builds the peanut-butter creature with your face and renders a 12-second 9:16 vertical clip on Seedance 2.0 — the skitter, the jar snatch, the pour, the spin, and the grocery-store rampage, with the high-pitched "give me peanut butter please" scream and street ambience as audio.

One selfie, a few minutes. No prompt writing, no model picking. Want a different rampage — a gas station instead of a grocery store, or a specific chant? Type it in Director's Notes — but keep your real face and the wet peanut-butter body, since that uncanny contrast is the whole joke.

The rest of this guide is for people who want to roll their own — change the creature, the rampage, or run it on a different model.

Or, Build It Yourself — What You Need

Three things:

1. One clear selfie of yourself. Front or three-quarter view, eyes open, good lighting.
2. An AI video model that accepts a reference image and generates audio. Seedance 2.0 is the safest — it preserves your face, handles the wet peanut-butter material and spindly-legged motion, and generates the screamed catchphrase in the same pass. A face-faithful reference model (Nano Banana Pro) helps lock your face onto the creature first.
3. A platform to post on. TikTok, Reels, and Shorts are where this travels.

You don't need any footage, props, or a single jar of peanut butter.

Step 1 — Pick Your Reference Photo

The selfie you feed the model is the face that ends up on the creature — and preserving it is the entire point. Choosing well saves wasted generations.

Use:

• A clear, well-lit photo of just you
• Front-facing or three-quarter angle, eyes open
• A head-and-shoulders shot (the model reads the face; the body becomes peanut butter anyway)

Avoid:

• Group photos (the model gets confused about whose face to keep)
• Sunglasses or anything covering the face (you lose the features that make it recognizably you)
• Low-resolution or motion-blurred shots
• AI-generated reference images (compounding artifacts)

Step 2 — Lock the Creature

The make-or-break of this format is the uncanny contrast: a calm, photoreal, unchanged human face sitting on top of an impossible body. Decide the details before you write the prompt:

• 😀 The face: your real face from the selfie, kept photoreal and completely unchanged — same skin, same features, calm expression. Never cartooned.
• 🥜 The body: made entirely of glistening, creamy, wet peanut butter — dripping, catching the light, clearly edible-looking goop, not clay and not matte.
• 🕷️ The limbs: long, thin, spindly insect / daddy-long-legs legs that let it skitter fast and low.
• 📱 The look: shaky vertical phone-cam footage, like a bystander caught it — an ordinary sidewalk and a normal grocery aisle, not a cinematic set.

Step 3 — Write the Rampage Prompt

This trend lives on one continuous rampage broken into timed beats, filmed like someone caught it on their phone. Feed the model your selfie as the reference and give it a time-segmented sequence. Copy this and swap in your specifics:

Keep the person's real human face from the reference photo exactly the same throughout — photoreal, unchanged, calm expression. Replace the entire body below the face with a creature made of glistening, creamy, wet peanut butter that drips and catches the light, held up on long, thin, spindly insect-like legs (daddy-long-legs). Shaky vertical phone-camera footage, like a bystander filming on the street — imperfect framing, handheld shake, autofocus hunting, ordinary real-world locations.[0 to 3s] A real city sidewalk: the peanut-butter creature skitters fast and low toward the camera on its spindly legs, screaming in a high-pitched voice.[3 to 6s] It snatches a peanut butter jar out of a passing stranger's grocery bag, twists it open, and pours the whole jar over its own head; the camera whip-pans to keep up.[6 to 8s] It spins in place, its legs smearing a thick peanut-butter ring across the pavement.[8 to 12s] It bursts into a grocery store and rampages the aisles, knocking jars off the shelves, then scrambles up a giant mountain of peanut butter.Audio: a high-pitched voice screaming "give me peanut butter please" on a loop throughout, plus real street ambience, startled bystanders, and clattering jars. No music track, so a trending sound can be added.Keep the human face consistent and photoreal in every beat; keep the peanut-butter body wet and glistening.

The non-negotiable elements:

• The preserved photoreal face — the calm, unchanged human face on the impossible body is the whole joke. State it up front and repeat it.
• The wet-glisten peanut butter — creamy, dripping, edible-looking from frame one. Matte or clay-like kills it.
• The spindly-legged skitter — the fast, low, insect-like movement is what makes it read as a creature, not a costume.
• The screamed catchphrase — "give me peanut butter please" on a loop does half the work. Keep "No music" so a trending sound can go over it when you post.

Step 4 — Pick a Model

• Seedance 2.0 — what Starrd runs on. Best at preserving your real face while transforming the body, handling the wet material and the spindly skitter, and generating the screamed catchphrase and ambience in the same pass. Safest pick.
• Nano Banana Pro — not the video model, but the best choice for locking your face onto the peanut-butter creature before you animate. Anchors hard to your reference face.
• Kling — strong on chaotic motion; workable, but push the "keep the real human face unchanged" language hard or it drifts the face.
• Veo 3.1 — capable with synced audio; add "shaky phone-camera, handheld, ordinary location" to fight its clean cinematic default.

No preference? Lock the creature on Nano Banana Pro, animate on Seedance 2.0. (See our Seedance vs Kling vs Veo breakdown.)

Step 5 — Generate and Iterate (Common Failure Modes)

First generations rarely nail it. The usual failures and fixes:

The face gets cartooned or drifts. This is the most common miss and it kills the joke. Use a clearer, front-facing selfie and state "keep the real human face photoreal and unchanged" explicitly. A face-faithful reference model helps lock it before you animate.

The peanut butter looks like clay or mud. Push the material language: "glistening, creamy, wet peanut butter that drips and catches the light." The wet-glisten from the first frame is what reads as peanut butter instead of a brown blob.

It looks like a polished VFX shot, not a phone clip. Add "shaky vertical phone-camera footage, handheld, autofocus hunting, ordinary sidewalk / normal grocery aisle." The found-footage realism against the impossible creature is the whole effect.

The rampage blurs into one motion. Keep the time segments tight and each beat a distinct action — skitter, snatch, pour, spin, rampage. Four or five clear beats in 12 seconds read as a rampage; a vague paragraph reads as mush.

The scream is missing or wrong. State the audio explicitly: a high-pitched voice screaming "give me peanut butter please" on a loop. The catchphrase is the signature — don't leave it to chance.

Budget 2 to 4 generations before a keeper — most of your iterating is on keeping the face photoreal.

Step 6 — Post It

Let the scream carry it — then add a trending sound. Keep the generated high-pitched catchphrase and the grocery-store chaos, and lay a trending audio over the top when you post (that's why the prompt says "No music").

Caption it as the creature, not a demo. "give me peanut butter please 🥜" or "he's just like me fr" travels further than "I made this with AI." (Within each platform's required AI-disclosure labels.)

Vertical, chaotic open. This is a 9:16 vertical format — open on the skitter toward the camera so the horror-comedy hits in the first 2 seconds, the scroll-stopper.

Common Mistakes That Tank Your Video

1. A cartooned or drifting face. The preserved photoreal human face is the whole joke. If it cartoons or morphs, re-roll and lock the face first.
2. Matte, clay-like peanut butter. Without the wet glisten it reads as a mud monster, not peanut butter. Push the creamy, dripping, light-catching material.
3. A cinematic camera. Clean, stabilized, cinematic framing kills the found-footage horror. Keep it shaky phone-cam in an ordinary location.
4. Dropping the catchphrase. "Give me peanut butter please" on a loop is the signature sound. Keep it, and add the trending sound over it.
5. Making it gross instead of funny. The tone is absurd horror-comedy, not actual body horror — keep the face calm and the chaos silly, so it reads funny.

Frequently Asked Questions

What is the trend? A person is turned into a creature that keeps their real face but has a body of glistening peanut butter and long spindly legs, skittering down the street and rampaging a grocery store while screaming "give me peanut butter please."

How do I make one? Preserve your face on a wet-peanut-butter creature with spindly legs on a face-faithful reference model, then animate a timed rampage with the screamed catchphrase on a native-audio model (Seedance 2.0) — or use the Give Me Peanut Butter Please template and just upload one selfie.

Why keep the real face? The uncanny contrast between a calm human face and an impossible peanut-butter body is what makes it funny and unsettling. Cartoon the face and the joke is gone.

How do I make it look real? Shaky phone-cam footage, wet-glisten peanut butter from the first frame, and an ordinary real-world location.

Do I need to label it as AI? Yes — all major platforms require it, and an obviously impossible creature should clearly be marked as AI anyway.

Related Reading

• How to Do the 'Generate a Man Next to Me That Would Suit Me' AI Trend — another AI-cast format: one selfie and AI casts a whole person around your face.
• How to Make a Mini Me AI Video — the tiny-person-in-hand trend: one selfie shrinks you into an adorable palm-sized version of yourself.
• How to Make an AI Squish Video — the glossy-material ASMR trend that turns your head into wet silicone, a cousin of the peanut-butter material effect.
• Viral AI Video Trends (2026): The Monthly Roundup — every trend worth making this month, each with a one-tap template.
• Seedance 2.0 Complete Guide — the model behind face-preserving creature transforms and in-pass audio.