Cofi Trend: Make the Coffee Making Fail AI Video

Two people stand at a kitchen counter. One scoops instant coffee next to the red mug, not into it. The other pours milk with total confidence, straight onto the counter. Nobody laughs, nobody speaks, and the camera never cuts. That is the Cofi trend, and if you have tried to recreate it you have probably hit the same wall as everyone else: the AI either fixes the mistake (the coffee lands neatly in the cup), breaks the faces, or turns the deadpan into a laughing stock-video couple.

This guide explains what the Cofi trend is, why the joke only works when the clip is played completely straight, and how to make your own version with Vogoo's image to video generator. It includes a clay (claymation) version, which is the safest way to do the "celebrity kitchen" variant, plus a photo version for couples and friends. The prompts below follow the same beat-by-beat structure we use in our Seedance 2.5 prompt guide, written for the video models available in Vogoo as of November 2026.

What is the Cofi trend?

The Cofi trend (also called the coffee making fail AI trend, "two people with zero synergy", or "when two legends try to make coffee") is a short vertical AI video in which two people try to make one cup of coffee together and get every step wrong. It began spreading on TikTok in late October 2026, and template versions in editing apps followed within days.

The format is very consistent:

• Setting: an ordinary home kitchen, a light stone counter, a red mug, a jar of instant coffee, a kettle or milk carton.
• Cast: two people side by side. Each is careful on their own; together they have zero coordination.
• Action: coffee powder goes beside the mug, water or milk is poured onto the bare counter, the mug gets moved at exactly the wrong moment.
• Tone: completely deadpan. No dialogue, no laughing, no reaction shots.
• Shape: one continuous shot, vertical 9:16, roughly 10–15 seconds.

Variants swap the cast: couples, roommates, coworkers, two dogs, or two famous-looking "legends". Some tool sites market that last version as a "celebrity kitchen AI" effect. The structure never changes, which is exactly why it is easy to recreate once you know the beats.

Why the coffee making fail is funny (and why AI keeps ruining it)

The humor is the gap between effort and result. Both people move slowly and seriously, as if they are doing something important, and the result is a counter covered in milk. The moment anyone smiles, it becomes a blooper reel instead of a deadpan joke.

That is also why most first attempts fail. Video models are trained mostly on footage where people succeed at what they are doing, so a vague prompt like "two people make coffee badly" usually produces one of three outcomes:

1. The model fixes the mistake. The powder lands in the mug, because that is what "making coffee" looks like.
2. The model adds emotion. Characters laugh, look at the camera, or react in shock, which kills the deadpan.
3. The objects drift. The mug changes color or position, hands merge with the kettle, and the counter rearranges itself mid-shot.

The fix is to describe the failure as the intended action ("pours milk directly onto the counter, missing the mug by a hand's width"), lock the emotion ("neutral expressions, no smiling"), and pin the props ("the red mug stays in the same spot").

Clay version vs photo version: which should you make?

Most Cofi trend guides only cover the photo version: upload two faces, get a realistic clip. That works for couples, but it is the wrong choice for the celebrity variant and it is the version most likely to go wrong. Here is how the two compare.

Our recommendation: make the clay version for anything involving famous-looking people. Real celebrities' faces in a fake video are a consent and platform-policy problem, while two clay "legends" (a stiff football captain and a pop diva in a sequin jacket, for example) get the same laugh without imitating anyone. Use the photo version only for people who are in on the joke.

How to make a Cofi trend video on Vogoo

You need two steps: a starting image, then image to video. The whole thing fits inside one generation.

1. Create or upload the starting frame. For the clay version, generate the still with text to image using the still prompt below. For the photo version, upload a clear, evenly lit photo of the two people standing at a kitchen counter (or generate one from their photos). Use 9:16.
2. Open image to video and upload the still as the first frame.
3. Pick a model and length. Seedance 2.5 and Kling 3 both handle multi-step physical action well; set the duration to 10–15 seconds and keep the aspect ratio at 9:16.
4. Paste one of the video prompts below, then generate. If a prop drifts or someone smiles, keep the image and regenerate with the fix from the troubleshooting table.
5. Want both characters to stay exactly on-model? Use reference to video, which takes several reference images plus a prompt, so you can feed each character separately. Our guide to keeping a consistent AI character across images covers how to build those reference sheets.

If you would rather not write prompts at all, browse Vogoo's video effects gallery for one-click trend templates and check which ones fit your clip.

Cofi trend prompts you can copy

Each prompt is written beat by beat, because video models follow a sequence far better than a summary. Change the bracketed parts and keep everything else.

Starting still: clay characters at the counter (text to image)

Claymation style still, stop-motion plasticine figures, two clay characters standing side by side at a light speckled stone kitchen counter, a glossy red mug in the center, a glass jar of instant coffee, a steel kettle and a milk carton. Left character: [a tall clay man in a tracksuit with a serious face]. Right character: [a short clay woman in a sequin jacket with a neutral face]. Wooden kitchen cabinets behind, soft daylight from a window, visible fingerprint texture on the clay, vertical 9:16, no text, no logos.

1. The classic coffee making fail (clay)

Claymation stop-motion style, one continuous static shot, vertical 9:16. Two clay characters stand at a kitchen counter with a red mug in the center. 0-4s: the right character carefully scoops instant coffee and pours it onto the counter next to the mug, not into it. 4-8s: the left character slowly picks up the milk carton and pours milk directly onto the bare counter, missing the mug by a hand's width. 8-12s: the right character moves the mug away just as the left character pours hot water where the mug used to be. 12-15s: both look down at the mess with completely neutral expressions. No smiling, no laughing, no talking, no camera cuts. The red mug keeps the same color and shape throughout.

2. Celebrity kitchen, the safe way (clay legends)

Claymation stop-motion style, single static shot, vertical 9:16. Two original clay characters styled as famous-looking "legends": a stiff football captain in a generic team jersey and a pop diva in a sequin jacket and sunglasses. They make coffee at a kitchen counter with total seriousness. The captain scoops coffee powder in a perfect line across the counter beside the red mug. The diva pours milk with a dramatic wrist flick, entirely onto the counter. The captain hands her the mug only after the milk carton is empty. Deadpan, no reactions, no words, no logos, no real people's faces.

3. Couple version from your photo (photo to video)

Keep both people's faces, hair, clothing and body proportions exactly as in the image. One continuous shot, vertical 9:16, natural kitchen daylight. The person on the right scoops instant coffee and drops it beside the red mug. The person on the left pours milk onto the counter instead of into the mug, slowly and confidently. The person on the right then moves the mug away just before water is poured. Both keep calm, neutral expressions the whole time. No smiling, no talking, no looking at the camera, no camera cuts, no extra people.

4. Two dogs making coffee (clay or realistic)

Claymation stop-motion style, single static shot, vertical 9:16. Two clay dogs, a golden retriever and a pug, stand on stools at a kitchen counter. The retriever holds a spoon in its mouth and drops coffee powder next to the red mug. The pug pushes the milk carton over with one paw, and milk spreads across the counter. The retriever carefully sets the empty spoon inside the dry mug. Both dogs stare forward, perfectly calm. No humans, no text, no camera movement.

5. Office break room edition

Realistic style, one continuous shot, vertical 9:16, office break room with a gray counter, a red mug and a coffee machine. Two coworkers in business casual clothes. One presses the machine button while the other is still holding the mug in the air, so coffee pours onto the counter. The second coworker places the mug under the spout only after the machine stops. Both nod seriously as if the job is done. Neutral faces, no laughing, no dialogue, no camera cuts.

6. Solo version (one person, zero synergy with themselves)

Claymation stop-motion style, single static shot, vertical 9:16. One clay character makes coffee alone at a kitchen counter with a red mug. They scoop coffee powder into the sugar bowl, pour milk into the coffee jar, then carefully stir the empty mug with great concentration. Completely deadpan, no smiling, no talking, no camera cuts.

Fixing the usual Cofi trend problems

What most Cofi trend guides skip

• The celebrity version is a likeness problem, not a prompt problem. Generating real stars' faces into a fake clip can break platform rules and the stars' rights. Clay "legends" keep the joke and drop the risk.
• Labeling depends on how realistic the clip is. TikTok asks creators to label AI-generated content that shows realistic-looking scenes or people, and YouTube's disclosure rules cover realistic altered or synthetic content but not clearly unrealistic animation. A clay clip is the easy case; a realistic photo version should carry the AI-generated label.
• The joke lives in the timing. Three slow beats beat six fast ones. If the generated clip feels frantic, the model is not the problem; the prompt has too many actions.

Posting your Cofi trend video

Keep the caption as dry as the clip: "we tried", "teamwork", "two legends, one mug". Post vertical, keep the original length, and turn on the platform's AI-generated label for any realistic version. For a series, keep the same two characters and change only the kitchen task: tea, cereal, pancakes. Recognizable characters are what make people come back for part two.

Ready to make yours? Start with a still in text to image, then bring it to life in Vogoo's image to video generator.

FAQ

What is the Cofi trend?

The Cofi trend is an AI video format in which two people calmly try to make one cup of coffee and get every step wrong, spilling powder and milk on the counter instead of the mug. It is shot as a single deadpan vertical clip of about 10–15 seconds.

Is the Cofi trend the same as the coffee making fail AI trend?

Yes. "Cofi trend", "coffee making fail AI", "two people with zero synergy" and "when two legends try to make coffee" all describe the same clip format.

Can I make the celebrity kitchen version with real celebrities?

We don't recommend it. Using real people's faces without their consent can break platform rules. Make original clay characters styled as famous-looking archetypes instead; the joke works the same way.

Do I need a reference video to make it?

No. A single starting image plus a beat-by-beat prompt in image to video is enough. A reference video or several reference images help only when you need tighter control over each character.

How long should a Cofi trend video be?

Around 10–15 seconds. That is enough for two mistakes and a final deadpan look; longer clips tend to drift.

Do I have to label it as AI-generated?

For realistic versions, yes on TikTok, which asks creators to label realistic AI-generated scenes or people. YouTube's disclosure covers realistic synthetic content but not clearly unrealistic animation, so a clay version is lower risk. Check each platform's current rules before posting.

Sources

• TikTok — Community Guidelines: Integrity and Authenticity: the requirement to label AI-generated or significantly edited content showing realistic-looking scenes or people.
• TikTok Support — About AI-generated content: how creators apply the AI-generated label when posting.
• YouTube Help — Disclosing use of altered or synthetic content: which realistic synthetic content must be disclosed, and why clearly unrealistic animation is exempt.

The trend description is based on publicly posted Cofi trend clips as of 2026-11-04. Prompt wording, the clay-first recommendation and the troubleshooting table are Vogoo's own. Platform rules change, so the sources above are the place to check current policy.