How to Make a Shoulder Carry AI Video: Lift, Flip, and Lean In
Two adults, one grey car park, about nine seconds. Seedream 5.0 locks the faces. The canvas holds the headlight. Seedance 2.5 only has to cross from the shoulders to the close-up.
This is a vertical video, not a lock screen, a phone background, or a desk setup. The two adults on this page were drawn with no photo attached. Do not try the lift for real. VibePaper does not ship a one-tap template, the TikTok sound, or anyone else’s face.
videosi, updated 14 November 2026, times the clip this way: the carry, a swing, a lean into the lens, then a thumbs-up, over about nine seconds. TrendTemplate’s how-to, updated 10 November, calls the common template about nine seconds and vertical 9:16. Seedance 2.5 will not cut a 2.5-second piece. In the model file the duration runs from 4 to 30 seconds, and Auto until you type a number.
On 16 November 2026, TrendTemplate’s weekly TikTok list put Shoulder Carry first. In the 24 hours before that check, 31 different accounts had posted an AI version. The main sound was picking up about 37,988 new videos a day. They mark the stage as Peaking, and the first time they saw it as 10 November 2026. Those counts are lower bounds from TikTok search samples, not totals. Their one-line description is the whole joke: carried on the shoulders, then you flip on top, “I’m the one”. The name “I’m the One” is the sound and the caption hook, not the name of the lift. videosi, published 10 November and updated 14 November, says TikTok often lists that audio as original sound. A template site has guessed a known producer. That link is unconfirmed, and this page does not repeat it as fact. Jump on the Trend, updated 16 November, places the widely remade clip in a grey underground car park: a fixed phone, vertical 9:16, about nine seconds. They estimate about 1.0 million views for the format, and they cite a TikTok by @777emilyyyy dated 24 October 2026, about 803,800 views and 41,700 likes as of that 16 November check. This guide does not round those figures up.
The person who was carried owns the close-up
The reports do not agree on the path. TrendTemplate’s list says you flip on top. videosi’s timing puts a swing between the carry and a close-up of the person now in front of the lens. One reading is up and over. The other is down in front. Both end in the same place: the adult who was on the shoulders is the one leaning in, then the thumbs-up. That is the punchline. This guide locks those two ends. It does not ask one prompt for both paths, because two directions in one sentence become a second flip, and a second flip is where extra arms appear.
Two adults, named in the prompt
Lock the faces before anyone is lifted
TrendTemplate’s how-to asks for one clear front-facing photo of the person who gets carried and ends in the close-up, plus an optional second photo for the carrier. On that template, roles follow upload order, not gender. VibePaper does not assign a role by upload order. You name the carrier and the carried person in the prompt, and you attach the portraits as references. Seedream 5.0 takes up to 14 reference images. Adults who agreed only. No celebrities, no public figures, no kids. If you have no photo you may use, generate original adults, as this page did. Even light, hair and shoulders in frame, no sunglasses. A later face swap has nothing clean to hold if the eyes are hidden.
The carrierStays standing. Faces the phone.
Carrier · Seedream 5.0
Front-facing photograph, head and shoulders. One adult, even soft light, hair fully in frame, no sunglasses, no hat, mouth closed, plain expression. Neutral grey wall. No text, no logo, no watermark. If a photo of an adult you may show is attached, keep that face and that hair. Do not imitate a celebrity or a public figure. This adult is the carrier: they stay standing and face the camera.
Paper Gen, before Generate. Set the ratio yourself. Attach a photo only of an adult who agreed. The face on this page was generated with no photo attached.
The one who is carriedEnds in the close-up. Not a gender.
Carried · Seedream 5.0
Front-facing photograph, head and shoulders, of a different adult. Even soft light, hair fully in frame, no sunglasses, no hat, mouth closed. Neutral grey wall. No text, no logo, no watermark. If a photo of an adult you may show is attached, keep that face and that hair. Do not imitate a celebrity or a public figure. This is the person who is carried and who ends in the close-up.
A different adult from the carrier. This is the face that has to survive the lean and the thumbs-up.
First frame. Carrier faces the phone. One leg kicked out.
The carry
Carry · Seedream 5.0
Vertical 9:16 photograph, full length. A grey underground car park at night. Cold concrete, one dull yellow floor line, a warm headlight raking dust across the bay. No readable signs, no logos, no plates. Camera locked at phone height, as if a phone is propped and not moving. One adult stands facing the lens, feet on the concrete. A second adult sits on their shoulders, legs to either side, one leg kicked out, both looking toward the lens, mouths closed, no sunglasses. Hair visible. Exactly two adults. No child. No third person. No text. No watermark. If reference photos are attached, the standing adult is the carrier and the adult on the shoulders is the one who ends in the close-up. Keep those faces, that hair, and those clothes.
Attach both portraits. Paper Gen, 9:16, before Generate. This still is the first frame of the video, not a tutorial for a real lift.
A body in mid-air is where the picture grows a spare arm. The sources also disagree on whether the carried person swings down in front or up on top, so a still of one path would pretend the other report is wrong. First / last frame on Seedance 2.5 is the tool for this beat: the carry goes in the first slot, the lean or the thumbs-up goes in the last, and the prompt only has to get from one to the other. The pictures on this page stop at the two ends.
03 4–7sThe lean
The carried person is now in front.
The lean
Lean · Seedream 5.0
Vertical 9:16. The same grey car park, the same headlight, the same dust, the same locked phone. The adult who was on the shoulders is now in front, close to the lens, leaning in. Shoulders and hair fill the frame. The carrier is a soft shape behind them. No sunglasses. Mouth closed. No text, no logo, no watermark. If a reference of that adult is attached, keep that face, that hair, and those clothes.
Attach the carried portrait, and the carry if the clothes were right. Same bay, same headlight. This can be the last frame if you are stopping before the thumbs-up.
Vertical 9:16. The same car park and the same light. The same adult who leaned in now gives one thumbs-up, still close, still owning the frame. One hand only, thumb up. The carrier stays behind, soft. Locked camera. No sunglasses. No text, no logo, no watermark.
Attach the lean if the face was right. One hand. If the thumb splits into two, regenerate the still. The video will not repair a hand.
One locked phone shot. No cut, no zoom, no camera move. Start with one adult facing the lens and a second adult seated on their shoulders, one leg kicked out. The adult on the shoulders leaves that perch and arrives in front of the lens, close, leaning in, then gives one thumbs-up. The standing adult remains behind. End on the thumbs-up. Exactly two adults. No extra limbs, no third person, no readable words, no logo. If sound is on, keep it to the room: a scuff on concrete and a short breath. Do not imitate a known song.
First / last frame. Carry in the first slot, thumbs-up in the last. Type 9 if 9 is in the range. The model file allows 4 to 30 seconds. Sound is your choice, and it is not the TikTok track. Camera movement off. Clear any style first. Aspect ratio becomes adaptive.
The shortest Seedance 2.5 clip is 4 seconds, so you still cannot match the 2.5-second carry on its own. Two clips, still First / last frame, still 9:16 stills. Edit them where you post. VibePaper does not assemble the template.
Carry to lean · Seedance 2.5
Locked phone, one shot, no zoom, no camera move. Start on the shoulder carry, one leg kicked out. The adult on the shoulders leaves that perch and arrives close to the lens, leaning in. End on the lean. Do not add a thumbs-up. Two adults only. No text, no logo.
First frame: the carry. Last frame: the lean. Duration 5. Do not ask this clip for the thumbs-up.
Locked phone, one shot, no zoom, no camera move. Start on the lean-in close-up. The same adult, already in front, gives one thumbs-up. End on that gesture. Two adults. No text, no logo.
First frame: the lean. Last frame: the thumbs-up. Duration 4 or 5. The gesture happens here, and only here.
The links fill the model and the prompt. They do not set Aspect ratio, Duration, the input mode, Sound, Draft preview, or a style. They do not upload a still or a photo. They do not spend credits and they do not press Generate. Paper Gen reads the prompt once, then removes it from the address bar. The longest prompt a link can hand over is 2,000 characters.
01
Draw the two faces in Paper Gen.
Image. Model Seedream 5.0, the default. Even light, hair and shoulders in frame, no sunglasses. Set the ratio before Generate. Attach a photo only of an adult you may show. “4 more like this” repeats the current prompt. Use it only when the face is close.
02
Draw the carry from both portraits.
Attach them as references. Grey underground car park, fixed phone, carrier facing the lens, the other adult on the shoulders, one leg kicked out. No readable sign. This is an AI reconstruction of a posted format. It is not a reason to attempt the lift.
03
Draw the lean and the thumbs-up.
Same bay, same clothes, same headlight. The carried person is now in front. Skip a mid-air flip still. First / last frame has two image slots, and the flip is the distance between them.
04
Grade the row, or grade nothing.
Continue in Canvas. Style lives in Style library and appends its own fragment at submit. Paper Gen has no style chip. For this bay, leave the library empty. The concrete and the headlight are already the grade. If you do grade, put the same preset on every still, then clear it before the video, so Seedance follows the graded picture instead of receiving the fragment again.
05
Move one shot in Seedance 2.5.
First / last frame. Carry in the first slot, thumbs-up in the last. That mode sets Aspect ratio to adaptive, so the stills should already be 9:16. Duration is Auto until you type a number. Type 9. Camera movement stays off. Push In dollies in. FPV Drone rushes the bay. This shot is a propped phone. Sound defaults to on in the model file and writes the model’s own audio, which is not the TikTok track. Reference mode can take an audio file you have the right to use, between 2 and 30 seconds. That is not lip-sync to a known song, and it is not a place for a track you do not own. Read the estimate on Generate. Draft preview, if you use it, locks that pass to 480p. The model file also offers 720p.
06
Replace the prompt if you used Make it a video.
Make it a video switches to video and drops the still into the first open reference slot. If the video prompt is empty, it copies the still prompt. A still prompt describes a pose, not the flip. Replace it, and switch the input mode to First / last frame yourself. The link on this page does not do that switch.
Leave the grade empty, unless every still shares it
The fragment is appended at submit. A preset on the video, after the stills are already graded, is a second grade.
Empty
The recommendation. The prompt already has cold concrete and one warm headlight. A preset on top of that is a second opinion about the light.
Showa Retro
Cinematic. Warm old film, a little grain, a quiet mood. It can sit with the headlight. Use it only if the carry, the lean, and the thumbs-up all get it. Clear it before the video.
Suspense
Cinematic. The fragment asks for a cold palette, low exposure, high contrast, and a thriller mood. It will bury the headlight. Leave it off unless you want a colder bay, and then only on every still.
Leave these off
Dark Gothic adds a crypt, faded grey, and candlelight. Cyberpunk adds neon. AI Realistic appends pore-level skin and a 35mm close-up, which fights a full-length bay. American Realistic asks for a Hollywood face and will argue with a reference photo. Pixar Character and Plush Toy turn the two adults into toys.
Whose face, and what you say when you post
The template sites ask for photos. VibePaper can take one as a reference. That does not make every photo fair to use, and it does not make the lift safe to try.
Adults who agreed
Your own photo, or another adult who said yes. Not a child. Not a screenshot of a template. Not a frame from someone else’s post.
No public figure
Do not attach a celebrity, a politician, or anyone the model would be imitating. If you have no photo you may use, generate an original adult, as this page did.
Do not perform the lift
The clip is an AI reconstruction of a format. A real shoulder carry is a fall waiting for a floor. This page is about pictures, not about trying it in a car park.
Say it is AI
Nothing in the clip was filmed. You may name the trend Shoulder Carry or “I’m the One”. Say the picture is generated. Add a sound only where you have the right to use it. VibePaper does not supply the TikTok track, and it does not lip-sync a copyrighted song.
The shoulders are the still. The close-up is the joke.
Seedream 5.0 for the two faces, the carry, the lean, and the thumbs-up. Style library empty, unless every still shares one preset and you clear it. Seedance 2.5, First / last frame, about nine seconds, camera locked. If a hand or a face is wrong, fix the still before you pay for the shot.
The 31 accounts, the 37,988 new videos a day, the Peaking stage, and the 10 November first-seen date are TrendTemplate’s 16 November 2026 reading of TikTok search samples. They are lower bounds, not totals. The nine-second, 9:16 template length is their how-to, updated 10 November. The four beats and the timing are videosi, published 10 November and updated 14 November. The car park, the fixed phone, the estimate of about 1.0 million views, and the @777emilyyyy post of 24 October 2026 (about 803,800 views and 41,700 likes as of 16 November) are Jump on the Trend, updated that day. This page does not average them into a new number. Search names people also use: “shoulder carry AI trend”, “I’m the one AI video”, “shoulder carry TikTok”.
No. It opens Paper Gen with Seedream 5.0 or Seedance 2.5 and the prompt filled in. It does not set the ratio, the duration, the input mode, Sound, or a style, and it does not press Generate. You spend credits only when you do.
Can I start from my photo?
Yes, if it is an adult who agreed. Attach it as a reference. Seedream 5.0 takes up to 14 reference images. Do not use a child, a celebrity, or a public figure. On a template site the carried role follows upload order. Here you name that role in the prompt. If you have no photo you may use, generate an original adult.
Why is there no picture of the flip?
The reports disagree on whether the carried person swings down in front or up on top. A still of one path would invent a winner. A body in mid-air is also where extra limbs show up. First / last frame gives Seedance 2.5 the carry and the close-up, and the prompt only has to cross the gap.
Can each beat be its own short clip?
Not at the posted lengths. videosi’s carry is about 2.5 seconds and the flip about 1.5. Seedance 2.5, in the model file, starts at 4 seconds and runs to 30. Type 9 for one shot, or split into a 5-second carry-to-lean and a 4- or 5-second lean-to-thumbs-up. Edit them where you post.
Is the sound the TikTok track?
No. “I’m the One” is the name of the sound and the caption hook, not a lyric this page will print. videosi says TikTok often lists the audio as original sound. A template site has guessed a known producer. That guess is unconfirmed. Seedance can write its own audio when Sound is on. That audio is not the track on the posts, and it is not lip-sync to a copyrighted song. Turn Sound off if you will add a sound you have the right to use in the app you post from.
Is Seedance 2.5 Flash in VibePaper?
No. Finish this clip on Seedance 2.5. Draft preview on that model locks the pass to 480p. The model file also offers 720p. Read the estimate before you generate. This page does not quote a price.
Why leave Style library empty?
Suspense appends cold light and a thriller mood, which fights the headlight. AI Realistic appends a close-up the bay is wider than. If you use a preset, Showa Retro is the one that can live with the tungsten, and only if every still gets it and you clear it before the video.