Copied iPhone Kupi prompts go wrong because the dance is a 15-second sequence with a phone that unfolds, flies through the air and comes back, and a pasted block of text controls none of that reliably. iPhone Kupi AI is an independent tool that builds the motion plan into one tap, so your video starts from a photo and a scene choice instead of a prompt you have to debug. This post explains the six faults people hit most and what actually fixes each one, without handing out a recipe.
The viral original is deceptively simple. In Win William's hotel-room clip, posted on 23 October 2026 and past 74 million views by 29 October, someone opens a book-style foldable phone toward the camera, steps back, dances with the open screen, tosses the phone, catches it and spins out of frame. Many viewers call it the iPhone Duo dance. It is seven small actions, one fixed low camera and no cuts, and each action is a place where a general video model can slip.
Key takeaways
- A prompt you find online is a snapshot of one lucky render, not a method. The same words give a different clip on the next run.
- Six faults show up again and again: a stranger's face leaking in, a foldable that unfolds wrong or grows, a phone that slows down or vanishes in the toss, a logo on the back, a face that drifts while dancing, and camera cuts the original never had.
- The fixes are almost never "add one more sentence". They are a better photo, a higher model tier or a fresh render.
- iPhone Kupi AI uses Seedance 2.0 / 2.5 from ByteDance, and you pick the tier. Likeness and motion are close, not guaranteed.
What is a pasted iPhone Kupi prompt actually asked to do?
A pasted iPhone Kupi prompt is asked to direct a whole 15-second routine in a few sentences, which is more than a general video model can follow exactly. One render has to deliver this, in order:
- Sit close to the lens with the folded phone near the chin (0 to 1 second).
- Push the half-open phone toward the camera, a screen full of apps (1 to 2 seconds).
- Throw the phone aside, lean in, make faces and sing along (2 to 4 seconds).
- Open the phone fully with both hands so a tablet-size screen fills the frame (4 to 5.5 seconds).
- Step back to full body and dance holding the open screen (5.5 to 9.5 seconds).
- Toss the phone up, out of frame, and catch it (9.5 to 10.5 seconds).
- Swing the phone, dance wildly, open the arms and walk out of frame (10.5 to 15 seconds).
The full beat sheet is on the iPhone Kupi trend page. What matters here is the shape: the camera never moves, the room never changes and the phone is a hinged object that has to change size believably. A prompt copied from a caption is usually a mood plus a few verbs. The video model reads it as a mood and fills the gaps itself, and the gaps are where the faults live.
Why does a stranger's face show up in my iPhone Kupi video?
A stranger's face shows up because many online recipes lean on a reference video of the original dancer, and a model that sees a reference can borrow from it instead of using only your photo. You might see Win William's features blend into yours for a second, or a first frame that looks like someone else and slowly becomes you.
This is a leak, not a styling choice. When a model gets both a photo and a clip, it has to decide which one owns the face, and in fast motion it can decide differently from frame to frame. Adding "keep my face" to the prompt does not settle it, because the model does not weigh that sentence against the reference the way a person would.
What fixes it: a different front-facing photo, a fresh render, or a higher tier if it keeps happening. iPhone Kupi AI sets the original up as a motion guide only and your photo as the face, so that part is not yours to manage. It still cannot promise a perfect likeness. The prompt page explains why no pasteable text is offered.
Why does the foldable unfold wrong or suddenly grow?
The foldable unfolds wrong because a hinge is hard to animate: the phone must open from a small slab into a tablet-size screen while the hands keep holding it. Typical versions of the fault are a phone that opens like a flip phone instead of a book, a screen that is already open at the start, a phone that changes size between frames, and a display bigger than the dancer's torso.
The original does it in stages. The phone opens halfway at 1 to 2 seconds and fully at 4 to 5.5 seconds. A pasted prompt usually says "opens a foldable phone" and nothing about stages, so the model picks one moment and sometimes skips the middle.
The size jump has a second cause. When the dancer steps back, the camera stays fixed, so the phone should keep its real size while the person shrinks in the frame. Models sometimes keep the phone the same size on screen instead, and it looks inflated.
The fix is a photo with the upper body and hands visible, a higher tier if the unfold keeps skipping, or a fresh render when the phone looks inflated.
Why does the phone slow down, float or disappear in the toss?
The phone slows down or disappears in the toss because the toss is the hardest second of the routine: a fast object leaves the frame and has to come back. Between 9.5 and 10.5 seconds in the original, the phone goes up, leaves the picture and is caught again. Models tend to handle that in one of three ways.
- Slow motion. The phone drifts up as if underwater and the dancer's motion softens to match, so the beat is lost.
- Vanishing. The phone leaves and never returns, and the hands stay empty for the rest of the dance.
- Duplicates. A second phone appears after the catch, or the phone changes color or shape on the way down.
A model has no physics engine. It predicts plausible next frames, and a fast object that leaves the picture gives it little to predict from. A longer sentence about throwing and catching adds nothing it can use.
Regenerate first, since the toss changes most between runs, then move up a tier if it fails twice. A calm scene from the scenes page also gives the model less to confuse with the phone.
Why does the phone grow an Apple logo?
The phone grows an Apple logo because the model has seen huge numbers of iPhone photos and often fills a blank phone back with the logo it expects, especially when the trend name says "iPhone". The back of a plain foldable is a flat dark surface, and the model decorates it.
The phone in the original is a book-style foldable, and whether it is an iPhone Duo is unverified, so the logo is not part of the dance. A made-up logo can also mislead viewers, which is a reason to avoid it.
iPhone Kupi AI is not affiliated with Apple. If a mark slips through, a fresh render is the fix, and a failed or blocked render returns its credits automatically.
Why does my face drift while I dance?
Your face drifts because the model has to carry your identity through a spin, a head turn and fast arm movement, and it loses a little each time. You see a jaw that changes shape, eyes that shift color, hair that changes length, or a face that looks right for 2 seconds and different by the end.
The riskiest part is the ending. From 10.5 to 15 seconds the dancer swings the phone, opens the arms and turns away, so the face is often at an angle or partly hidden and the model invents more.
What you control before pressing the button matters most:
- Photo angle. A front-facing photo drifts less than a three-quarter one.
- Light. Even light on the whole face beats strong side shadows.
- Obstructions. Sunglasses, hats, hands over the face and heavy filters all make drift worse.
- Tier and resolution. Higher tiers and 1080p keep detail longer.
The how to make an iPhone Kupi video page covers the photo basics. Adding "do not change my face" to a prompt replaces none of it.
Why do extra camera cuts appear in my iPhone Kupi video?
Extra cuts appear because video models are trained on edited footage and like to cut: wide shot, close-up, wide shot. The original has none. It is one fixed low camera in a hotel room for all 15 seconds, and that stillness is part of why it looks like a real phone video.
A prompt that says "dance video" or "viral TikTok" invites cuts, because most dance videos in the training data have them. You might get a close-up of the phone screen mid-unfold or a new angle on the toss, and the continuity of the dance is gone. Telling a model "no cuts" is a request, not a rule, so some runs obey and some do not. Render again, choose an uncluttered scene, or trim the end if only the last second misbehaves.
What fixes each fault?
Each fault has one or two fixes that work, and none of them is more text.
| Fault | What you see | Change this first | If it still fails |
|---|---|---|---|
| Face leak | Another person's features blend in | New front-facing photo | Render again, then a higher tier |
| Foldable unfolds wrong | Opens the wrong way or grows | Photo with hands and torso visible | Higher tier |
| Toss goes wrong | Slow motion, phone vanishes | Render again | Higher tier, calmer scene |
| Logo on the phone | A mark appears on the back | Render again | Different scene |
| Face drift | Face changes while dancing | Sharper, even-light photo | Higher tier or 1080p |
| Extra cuts | Angle changes mid-clip | Render again | Calmer scene, or trim the end |
iPhone Kupi AI offers Standard (Seedance 2.0 Mini), Fast (Seedance 2.0 Fast), Pro (Seedance 2.0) and Ultra (Seedance 2.5). A Standard 720p video costs 10 credits, and the range across tiers and resolutions runs from 5 to 93 credits. The pricing page has the packs. Failed or blocked renders return their credits automatically, but a finished video you dislike has used its credits, so fix the photo before paying for a second render.
What does iPhone Kupi AI do differently?
iPhone Kupi AI builds the dance structure in, so your only inputs are a photo and a scene. The 15-second beat sheet is part of the render, the phone is set as a generic foldable with no logo, the camera is fixed, and the original sound is added to the finished video afterward because the model renders silent.
You pick from 7 scenes, including the original hotel night and a two-person version on the iPhone Duo dance page. The result is an MP4 in 9:16 by default, with 1:1 and 16:9 also available, downloadable for 24 hours. None of this makes the faults vanish. It moves the work from writing text to choosing a good photo and a tier, a simpler job you can judge by looking.
What are the limits?
- Likeness and motion are close, not exact. A finished video can still show a drifting face or a failed toss.
- There is no free tier, because each video runs on a paid model, and credit packs are not refundable.
- iPhone Kupi AI is unofficial and not affiliated with Apple, TikTok, Win William, SABI or MIA BOYKA.
- Upload only photos of adults who agreed to be in the video, never celebrities.
- The song is not licensed for ads, so keep these videos organic.
When you are ready, make your video from one clear photo and judge the first result before trying another tier. For the lyric and the song, see the iPhone Kupi song page.
Frequently asked questions
Why does my iPhone Kupi video show someone else's face for a moment?
Your iPhone Kupi video can flash another face because the model sometimes borrows from the motion reference instead of only your photo. It is usually a one-off. On iPhone Kupi AI, render again, or upload a sharper front-facing photo with even light. A failed render returns its credits automatically, but a finished one uses 10 credits at Standard 720p.
Should I move up a model tier when the phone keeps disappearing?
Yes, move up a tier if the toss fails on two renders in a row, because stronger tiers track an object that leaves the frame more reliably. iPhone Kupi AI offers Standard, Fast, Pro and Ultra, built on Seedance 2.0 and 2.5, from 5 to 93 credits depending on tier and resolution. Try one fresh render at the same tier first.
Will a prompt that worked for someone else work for me?
No, a prompt that worked for another person will not reliably work for you. The same words produce a different render each time, and the photo changes the result more than the wording does. iPhone Kupi AI does not publish a prompt because the dance timing is built in, and you only choose a photo, a scene and a tier.
Can I fix a glitch in the last second without paying for a new render?
Often you can, if the first 13 or 14 seconds are right. Download the iPhone Kupi AI MP4 within 24 hours and trim the end in any editor or in the app you post from. This works for a stray cut or a face that drifts at the very end, but faults early in the clip need a fresh render.
Why does my iPhone Kupi video have no brand logo on the phone?
The phone in an iPhone Kupi AI video is a generic book-style foldable with no logo, on purpose. The site is not affiliated with Apple, and whether the original phone is an iPhone Duo is unverified. If a logo-like mark appears, that is a generation slip, and a new render usually removes it.