The Helping Hand
Find an accomplice
Prompt
SEEDANCE 2.5 · VIDEO PROMPT · 15S · 9:16 FORMAT: 15s / 9:16 vertical / viral handheld comedy / one continuous shot, no cuts, no zoom STYLE: Ultra realistic smartphone handheld footage, natural overcast daylight, real world physics, slight handheld shake from a friend filming, no cinematic grading, no slow motion, documentary authenticity. REFERENCES: @Image 1 defines the environment, camera vantage and wardrobe ONLY: a tall old brick and stone wall on a cliff edge above a hazy canyon, a thick natural fiber rope hanging over the top edge, and two outfits. Outfit A (man on top of the wall): blue satin varsity jacket with orange-and-white striped cuffs and embroidered patches, dark navy baseball cap. Outfit B (climbing man): burgundy wool varsity jacket with cream striped trim and chenille letter patches, dark denim jeans, grey suede sneakers, black baseball cap. Use its framing, wall texture, rope, cliff and haze exactly. Do NOT use its faces. VIDEO PROMPT: [Stage 1 — approx 0–4s] Initial state: the handheld phone camera holds the same vantage as @Image 1, slightly below the top of the wall, looking at the brick face and the canyon haze behind it. Only ONE man is visible: the blue-jacket man leans over the top edge of the wall gripping the rope with both hands. The rope runs taut straight down the wall and exits the bottom of frame. Nobody else is in frame. Event: he pulls hand over hand with visible effort; his shoulders strain, his cap tilts, small dry dust falls from the bricks where the rope rubs. Camera: restrained handheld, small breathing shake, no pan. End state: the blue-jacket man is still on top gripping the rope, the rope is taut, the lower half of the wall is empty. [Stage 2 — approx 4–9s] Continue from that state. Event: the burgundy-jacket man rises slowly into frame from the bottom edge, climbing the rope, his grey sneakers planted flat against the bricks, body leaning back away from the wall, cap visible, sunglasses on, face tilted up toward the top. He gains roughly one brick row per pull, slow and effortful, nothing sudden. Camera: same handheld hold, follows nothing, keeps both men in frame. End state: the burgundy-jacket man is about one arm length below the top edge, both hands on the rope, feet on the wall; the blue-jacket man still holds the rope from above. [Stage 3 — approx 9–12s] Continue from that state, everything at normal real-time speed. At 9s the blue-jacket man releases the rope with his right hand and reaches down toward the climbing man with an open palm, keeping his left hand on the rope. The burgundy-jacket man glances up, lets go of the rope with his right hand and snaps his hand up to the offered palm in one quick natural motion; the grip lands immediately, no pause, no stretched moment, the frame rate and motion speed stay exactly as in the previous stages. Right after the grip the burgundy-jacket man's feet scrape and slip off the bricks. End state: right hands locked together, the blue-jacket man bent far over the edge, the burgundy-jacket man hanging from that grip with feet off the wall. [Stage 4 — approx 12–14s] Continue from that state. Event: the added weight pulls the blue-jacket man forward; his chest slides over the top edge of the wall and he tips over it, still holding the other man's hand. Both men drop together straight down along the wall face and exit the bottom edge of frame within one second, the rope whipping after them. The camera does NOT tilt down to follow them. End state: no people in frame. [Stage 5 — approx 14–15s] Hold the empty composition: only the brick wall, the loose rope end swinging against the bricks, dust drifting down, canyon haze behind. Hard stop on the empty frame. AUDIO: <light wind across the cliff, distant canyon ambience, rope fibers creaking under load, grit and small stones scraping on brick, strained breathing and effort grunts from the blue-jacket man, sneaker rubber scuffing brick as the second man climbs, a short surprised shout from both men as they tip, then only wind and the rope tapping the wall> No music, no voiceover, no dialogue. CONSISTENCY: Exactly two men, both with the identical face from @Image 2. Outfit A stays on the man on top, Outfit B stays on the climbing man; no wardrobe swap. Sunglasses and caps stay on both men in every stage. One single rope. Camera vantage, wall, cliff and haze unchanged from @Image 1. Camera never pans, tilts or zooms. Constant real-time playback speed from first frame to last; the hand grab plays at the same speed as the climbing. The burgundy-jacket man is absent until Stage 2. Both men exit the bottom of frame together in Stage 4 and never reappear. NEGATIVE PROMPT: no third person, no duplicate hands, no merged hands, no face change between the two men, no different faces, no wardrobe change, no camera pan or tilt, no slow motion, no time ramp, no speed change at the hand grab, no cuts, no cinematic color grade, no floating, no impact shown, no landing shown, no on-screen text, no captions, no watermark, no music.
Tutorial
DENIZ AKKABAK · SEEDANCE 2.5 TUTORIAL · @DENIZ.AKKABAK
MADE IN INVIDEO · invideo.io/i/DENIZ-ig

The Helping Hand
One face, two jackets, one rope — how a Nano Banana Pro still and a five-stage Seedance 2.5 prompt turn a rescue into a punchline.
A TUTORIAL BY DENIZ AKKABAK · @DENIZ.AKKABAK
PIPELINE NANO BANANA PRO → SEEDANCE 2.5 · FORMAT 15S · 9:16 · GENRE HANDHELD COMEDY
Two men on a cliff wall. One is on top, hauling a rope; the other is climbing it. The top man reaches down, the climber grabs the hand — and the extra weight tips the rescuer over the edge, so both of them drop out of frame together. Then the camera just stays there, looking at an empty wall and a swinging rope. That is the whole video, and it works because it is shot the way a friend would actually shoot it: one take, no zoom, no slow motion, no reaction cut. The gag is delivered by physics, not by editing.
The trick underneath it is that both men are the same person. A single breakfast-table photo becomes the identity reference for a photoreal Nano Banana Pro still in which that one face appears twice, in two different varsity jackets. That still then becomes the only reference Seedance 2.5 sees — the face, the wardrobe, the wall, the rope and the camera vantage all travel in one image, so the video prompt never has to describe a person again. It only has to describe what happens to him.
Below is the exact two-step pipeline — both prompts verbatim, both run inside invideo — followed by an anatomy of why the fall lands.
THE PIPELINE AT A GLANCE
- 01
Nano Banana Pro — clone yourself onto the wall
One identity photo in, one photoreal 9:16 still out: the same man on top of the wall in a blue satin bomber and far below on the rope in a cherry-red varsity jacket. Everything the video needs — face, outfits, wall, rope, vantage — is fixed here.
- 02
Seedance 2.5 — one reference, five stages
The still is attached as @Image 1 and a 15-second prompt walks the model through five timed stages, each with an explicit end state, from a quiet rope pull to an empty frame.
STEP 01 — NANO BANANA PRO
Clone Yourself Onto the Wall
REFERENCE INPUT — IMAGE 1, IDENTITY

The still has to do three jobs at once: put one real face on two bodies, dress those bodies in two clearly different outfits, and frame the wall exactly the way the video will later be framed. So the prompt opens with the reference contract before a single word of scene description — Image 1 is the face, hair, skin, facial hair and eyewear, and both men are that person. A second image for the red jacket is declared optional; the prompt describes the jacket precisely enough that you can skip it, which is what happened here.
The scene itself is written like a camera report, not a mood board: iPhone 15 Pro, 26mm, f/1.8, vertical, the phone held slightly below the top edge of the wall and tilted a little upward from a ledge across the gorge. That vantage line is the same one the Seedance prompt will quote back later, so the still and the video are already speaking the same language. Note the small honesty cues — a small soft lens flare in the top-left corner, slight computational sharpening, in-between expressions of effort — that are what make a generated image read as a phone snapshot rather than a render.
PROMPT — COPY EXACTLY
| NANO BANANA PRO · IMAGE PROMPT · 9:16 |
|---|
| Image 1: identity reference — use this person's face, hair, skin tone, facial hair and any eyewear exactly. Both men in the scene are this same person. Image 2 (optional): reference for the red jacket. |
| Amateur smartphone photo, candid, shot on an iPhone 15 Pro 26mm f/1.8, vertical 9:16. The camera is on a rocky ledge directly across a gorge from a tall old brick wall, held slightly below the top edge of the wall and tilted a little upward, so the wall's face fills most of the frame, its top edge cuts across the upper third, and the wall's right corner drops away to reveal hazy canyon cliffs and distant treeline behind it. Bright overcast sky above, a small soft lens flare in the top-left corner. |
| The wall: weathered brick in cobblestone tones — warm greys, ochre, faded terracotta, patches of rough stone — mortar eroded, older rubble stone at the base, running straight down out of frame with no ground visible. |
| Top man: the person from Image 1 leans over the top of the wall from behind it, chest on the parapet, looking straight down. He wears a royal-blue nylon satin bomber jacket covered in orange, white and blue embroidered basketball-team patches and tackle-twill appliqué on the chest and sleeves, orange-white striped ribbed collar and cuffs, snap front; a navy structured snapback cap with embroidered front artwork. Both hands clamp a coarse jute rope, elbows bent, forearms tensed, visibly hauling. |
| Lower man: the same person, far below in the bottom third of the frame, noticeably smaller for distance, gripping the rope with both hands above his head, sneakers braced flat against the brick, one knee bent, body leaning back from the wall, face tilted up toward the man above. He wears a deep cherry-red microsuede varsity jacket with a soft napped texture, ribbed red collar, cuffs and hem with a thin cream stripe, snap front, embroidered chenille patches — a gothic "K" on the left chest, a small "11" on one sleeve, a round crest on the opposite shoulder, cream and navy thread. Underneath a white tee, dark indigo jeans, grey suede-and-mesh running sneakers with a white sole, and a black low-profile cap. |
| The rope is one taut straight line from the top man's hands to the lower man's grip, fibers strained, a knotted loop under the lower hands, a faint scuff of brick dust where the sneakers touch the wall. Soft shadows on the brick consistent with the overcast top light. |
| Crisp sharp focus across the entire frame, realistic colors, clean HDR look, slight computational sharpening, natural skin texture on both faces, in-between expressions of effort. No filters, no text, no watermark, no extra people, no deformed hands. |
OUTPUT — CHECK IT BEFORE YOU MOVE ON

This single image is the next step's entire reference set, so audit it like one. The two faces should read as the same man — same curls under the caps, same sunglasses, same mustache. Blue must be on top and red on the rope, because the video prompt will name the characters by jacket. The rope should be one taut line from hands to hands, the wall's right corner should drop into canyon haze, and the sneakers should sit flat on the brick with a scuff of dust. If any of that drifted, regenerate now — it is far cheaper to fix a still than a 15-second video.
STEP 02 — SEEDANCE 2.5
One Reference, Five Stages
REFERENCE INPUT — THE ONLY IMAGE ATTACHED

Attach the still as @Image 1 — nothing else. The reference paragraph then tells Seedance what to take from it: the wall on its cliff, the rope over the edge, the camera vantage, and two named outfits. Outfit A is the man on top, Outfit B is the climber. Naming the outfits is what lets the rest of the prompt say the blue-jacket man and the burgundy-jacket man without ever re-describing them — every stage refers to the two characters by wardrobe, and the wardrobe is pinned to the image.
The video prompt is five stages on a timeline. Each one starts by inheriting the previous stage's end state and ends by declaring its own, so the model always knows where the last beat left the scene. Stage 1 is a single man and a taut rope. Stage 2 brings the climber up into frame roughly one brick row per pull. Stage 3 is the hand grab, written with an almost paranoid insistence on real-time speed. Stage 4 is the tip and the drop, exiting the bottom edge within one second, camera not following. Stage 5 holds the empty wall. Set the model to 15 seconds and 9:16 to match the FORMAT line.
PROMPT — COPY EXACTLY
| SEEDANCE 2.5 · VIDEO PROMPT · 15S · 9:16 |
|---|
| FORMAT: 15s / 9:16 vertical / viral handheld comedy / one continuous shot, no cuts, no zoom |
| STYLE: Ultra realistic smartphone handheld footage, natural overcast daylight, real world physics, slight handheld shake from a friend filming, no cinematic grading, no slow motion, documentary authenticity. |
| REFERENCES: |
| @Image 1 defines the environment, camera vantage and wardrobe ONLY: a tall old brick and stone wall on a cliff edge above a hazy canyon, a thick natural fiber rope hanging over the top edge, and two outfits. Outfit A (man on top of the wall): blue satin varsity jacket with orange-and-white striped cuffs and embroidered patches, dark navy baseball cap. Outfit B (climbing man): burgundy wool varsity jacket with cream striped trim and chenille letter patches, dark denim jeans, grey suede sneakers, black baseball cap. Use its framing, wall texture, rope, cliff and haze exactly. Do NOT use its faces. |
| VIDEO PROMPT: |
| [Stage 1 — approx 0–4s] Initial state: the handheld phone camera holds the same vantage as @Image 1, slightly below the top of the wall, looking at the brick face and the canyon haze behind it. Only ONE man is visible: the blue-jacket man leans over the top edge of the wall gripping the rope with both hands. The rope runs taut straight down the wall and exits the bottom of frame. Nobody else is in frame. Event: he pulls hand over hand with visible effort; his shoulders strain, his cap tilts, small dry dust falls from the bricks where the rope rubs. Camera: restrained handheld, small breathing shake, no pan. End state: the blue-jacket man is still on top gripping the rope, the rope is taut, the lower half of the wall is empty. |
| [Stage 2 — approx 4–9s] Continue from that state. Event: the burgundy-jacket man rises slowly into frame from the bottom edge, climbing the rope, his grey sneakers planted flat against the bricks, body leaning back away from the wall, cap visible, sunglasses on, face tilted up toward the top. He gains roughly one brick row per pull, slow and effortful, nothing sudden. Camera: same handheld hold, follows nothing, keeps both men in frame. End state: the burgundy-jacket man is about one arm length below the top edge, both hands on the rope, feet on the wall; the blue-jacket man still holds the rope from above. |
| [Stage 3 — approx 9–12s] Continue from that state, everything at normal real-time speed. At 9s the blue-jacket man releases the rope with his right hand and reaches down toward the climbing man with an open palm, keeping his left hand on the rope. The burgundy-jacket man glances up, lets go of the rope with his right hand and snaps his hand up to the offered palm in one quick natural motion; the grip lands immediately, no pause, no stretched moment, the frame rate and motion speed stay exactly as in the previous stages. Right after the grip the burgundy-jacket man's feet scrape and slip off the bricks. End state: right hands locked together, the blue-jacket man bent far over the edge, the burgundy-jacket man hanging from that grip with feet off the wall. |
| [Stage 4 — approx 12–14s] Continue from that state. Event: the added weight pulls the blue-jacket man forward; his chest slides over the top edge of the wall and he tips over it, still holding the other man's hand. Both men drop together straight down along the wall face and exit the bottom edge of frame within one second, the rope whipping after them. The camera does NOT tilt down to follow them. End state: no people in frame. |
| [Stage 5 — approx 14–15s] Hold the empty composition: only the brick wall, the loose rope end swinging against the bricks, dust drifting down, canyon haze behind. Hard stop on the empty frame. |
| AUDIO: <light wind across the cliff, distant canyon ambience, rope fibers creaking under load, grit and small stones scraping on brick, strained breathing and effort grunts from the blue-jacket man, sneaker rubber scuffing brick as the second man climbs, a short surprised shout from both men as they tip, then only wind and the rope tapping the wall> No music, no voiceover, no dialogue. |
| CONSISTENCY: Exactly two men, both with the identical face from @Image 2. Outfit A stays on the man on top, Outfit B stays on the climbing man; no wardrobe swap. Sunglasses and caps stay on both men in every stage. One single rope. Camera vantage, wall, cliff and haze unchanged from @Image 1. Camera never pans, tilts or zooms. Constant real-time playback speed from first frame to last; the hand grab plays at the same speed as the climbing. The burgundy-jacket man is absent until Stage 2. Both men exit the bottom of frame together in Stage 4 and never reappear. |
| NEGATIVE PROMPT: no third person, no duplicate hands, no merged hands, no face change between the two men, no different faces, no wardrobe change, no camera pan or tilt, no slow motion, no time ramp, no speed change at the hand grab, no cuts, no cinematic color grade, no floating, no impact shown, no landing shown, no on-screen text, no captions, no watermark, no music. |
OUTPUT

STEP 03
Anatomy — Why the Fall Lands
Seven decisions in the Seedance prompt do most of the work. Each is tied to a specific line you can find in the box above.
- 01
The contract
The FORMAT line is a set of promises the model has to keep: 15s / 9:16 vertical / one continuous shot, no cuts, no zoom. A comedy of this kind dies the moment the model inserts a reaction shot, so the no-cut rule is stated first, before any story.
- 02
Realism as a style
STYLE asks for ultra realistic smartphone handheld footage with slight handheld shake from a friend filming and explicitly no cinematic grading, no slow motion. The joke needs to look overheard, not staged — the prompt refuses every tool that would make it look expensive.
- 03
Identity locked upstream
The face was solved in Step 01, so the reference paragraph can spend all of its words on environment, camera vantage and wardrobe: the wall, the rope, and Outfit A versus Outfit B. Every later line addresses the characters as the blue-jacket man and the burgundy-jacket man — wardrobe is the name.
- 04
Stages that hand off
Every stage closes with End state: and the next opens with Continue from that state. Stage 1 even declares that the lower half of the wall is empty — the model is told what is not there, so the climber's entrance in Stage 2 is an event rather than a continuity error.
- 05
The grab at real speed
Stage 3 is the risk point: video models love to dramatise a hand reaching for a hand. The prompt shuts that down line by line — in one quick natural motion; the grip lands immediately, no pause, no stretched moment, the frame rate and motion speed stay exactly as in the previous stages. The same demand is repeated in CONSISTENCY and again in the negatives.
- 06
Sound as structure
The AUDIO bracket is the timeline in another medium: rope creaking, grit on brick, effort grunts, sneaker rubber, a short surprised shout from both men as they tip, then only wind and the rope tapping the wall. No music, no voiceover, no dialogue keeps it inside the phone-footage contract.
- 07
The insurance
The NEGATIVE PROMPT protects the ending as much as the faces: no impact shown, no landing shown keeps the fall implied, no camera pan or tilt stops the phone from following them down, and no third person, no duplicate hands, no merged hands guard the two-body grip that the whole gag hangs on.
THE RESULT
Output — The Helping Hand

Check it against the prompt: one man for the first four seconds; the climber rising from the bottom edge and gaining a row at a time; the right hand offered and taken at the same speed as everything else; the tip, the drop, the rope whipping after them; and then a second of empty wall, dust drifting, haze behind. The camera never moves. Nothing lands. That is what a five-stage prompt with honest end states buys you — a gag that plays entirely inside a single frame.
THE JOKE IS WRITTEN IN THE END STATES
Lock the face once. Then let the physics tell the joke.
Swap the rope for a ladder, the wall for a balcony, the varsity jackets for whatever hangs in your wardrobe — the skeleton carries any two-person gag. One identity photo becomes a photoreal still in Nano Banana Pro; that still becomes the only reference Seedance 2.5 needs; five stages with explicit end states walk the model from a quiet setup to an empty frame. The camera never moves, the speed never changes, the fall is never shown — that restraint is the punchline.
WRITTEN, DIRECTED & PROMPTED BY DENIZ AKKABAK · @DENIZ.AKKABAK

Both models in this guide run inside invideo — Nano Banana Pro for the still, Seedance 2.5 for the video — with the reference passed straight from one step to the next. To rebuild this video or shoot your own version of the gag, start here: