Clone one character into the same scene three or four times in GPT Image 2.0, then drop the still into Seedance 2.0 — the closest version of you locks frozen like a statue while every other copy comes to life: one drives past in a car, one rides a bike, one sets the table. A single subject, multiplied, half-frozen, half-alive.
A two-tool workflow that turns one character photo into a surreal "clones in one frame" scene. First, GPT Image 2.0 duplicates your subject several times inside a single cinematic shot — the same person standing center, sitting in a car, riding a bike, prepping a table — all sharing one identity, one outfit, one face. Then Seedance 2.0 animates it: the foreground copy stays perfectly frozen like a paused photograph while every other clone and the entire world behind them moves naturally. The illusion lives in the contrast between the one body that's stopped and the four that are alive.
The Workflow
FIVE STEPS
01
Setup · Open The Model
GO TO GPT IMAGE 2.0
Open a platform with access to GPT Image 2.0. This is the image stage — it's the model that will take a single photo of your character and clone them into the same frame multiple times while keeping the face, outfit, and body consistent across every copy. Start a fresh image generation with image input enabled so you can attach your reference.
Open GPT Image 2.0
Start a fresh generation
Make sure image input is on
02
Input · Your Character
UPLOAD THE CHARACTER IMAGE
Upload one clear photo of the person you want to multiply. A sharp, well-lit shot with the full face visible works best — that's what every clone in the scene inherits. If your scene calls for an accessory (like the bucket hat in Scene B & C below), have that design image ready to attach too, so GPT can dress each copy in it.
One clear character photo
Full face, good lighting
Optional: bucket hat / accessory image
Attach as the reference
03
The Brief · Duplicate Prompt
PASTE THE SCENE PROMPT
Paste one of the three scene prompts below into GPT Image 2.0 — or write any similar scene of your own. The recipe is always the same: put the same character in several roles inside one wide cinematic shot, ask for a 4:5 ratio, and end with "maintain the character image" so every clone keeps the original face. Pick a vibe and copy it.
Scene A — Dubai Street
make this guy wearing sunglasses, he's standing in the middle of a dubai street, behind him is the same character in a car looking at him from the window, the same character is also on a bike from the other side, and the same character is preparing the table, wide cinematic high quality shot, 4:5 ratio, maintain the character image
Scene B — BBQ Grill
make this guy wearing sunglasses, make him wear the attached bucket hat design, he's next to a bbq grill with food on it, the same character is also behind on the phone, and the same character is preparing the table, wide cinematic high quality shot, 4:5 ratio, maintain the character image
Scene C — Pool Jump
make this guy wearing sunglasses, make him wear the attached bucket hat design, he's wearing shorts and white tshirt and jumping in the pool, behind him is the same character as life guard surprised, and on the side is the same character as a waiter looking also at him, wide cinematic high quality shot, 4:5 ratio, maintain the character image
Build your own: keep the formula — one "hero" copy in front + 2–3 more copies of the same character doing different things around the scene + wide cinematic high quality shot, 4:5 ratio, maintain the character image.
04
Handoff · Switch Tools
DOWNLOAD & GO TO SEEDANCE 2.0
Once GPT Image 2.0 gives you a clean frame with all the clones placed naturally, download the image. Then open Seedance 2.0 and start a new image-to-video generation, uploading that downloaded still as your reference. This is where the still becomes a living shot — Seedance freezes the front copy and animates the rest.
Download the GPT image
Open Seedance 2.0
Start image-to-video
Upload the still as reference
05
The Animation · Full Prompt
PASTE THE SEEDANCE PROMPT
Paste the full prompt below into Seedance 2.0 exactly as written and generate. It does two jobs at once: it freezes the foreground copy completely — no blink, no breath, no micro-movement, like time stopped only for them — while every background clone and the whole environment animates independently with believable, varied human motion. The camera stays locked off. When it's done, download and enjoy :)
Seedance 2.0 Prompt — Full Brief
Input: Use the attached image as the absolute visual reference. Preserve every character exactly as shown, including facial identity, hairstyle, clothing, accessories, body proportions, poses, lighting, environment, composition, and perspective. Do not redesign, replace, or reinterpret any subject. The camera framing must remain identical to the reference image.
Prompt:
Create a premium cinematic animation from the attached still image using a completely static locked-off camera. The framing, focal length, perspective, and composition never change throughout the shot. No camera movement, zoom, pan, tilt, dolly, handheld motion, or reframing.
The primary foreground character closest to the camera remains perfectly frozen for the entire duration, as if time has stopped only for them. Their body, face, head, eyes, hands, clothing, hair, and accessories remain absolutely motionless with no blinking, breathing, shifting, or micro-movements. They should appear like a perfectly still photograph.
Everything behind the foreground character comes to life naturally in real time. All background characters animate independently with realistic human behavior appropriate to the scene. They walk, talk, gesture, turn their heads, interact with each other, laugh, sit down, stand up, use objects, and continue their activities with believable timing and varied motion. Every character should move differently to avoid synchronized animation.
The environment is also alive with subtle natural motion: clothing reacts to movement, hair responds to wind, shadows shift naturally, reflections update correctly, leaves move gently, signs or decorations sway slightly, dust particles drift through the air, ambient lighting flickers subtly where appropriate, and any vehicles or environmental elements continue moving realistically.
The frozen foreground subject acts as a visual anchor while the rest of the scene evolves around them, creating a striking "time has stopped for one person" effect. Preserve realistic physics, accurate depth relationships, occlusion, lighting consistency, and motion blur only on moving elements. The frozen character remains perfectly sharp at all times.
Maintain a high-end cinematic look with AAA-quality realism, physically accurate rendering, smooth natural animation, and seamless continuity. No surreal effects, morphing, warping, camera shake, object deformation, or changes to the original scene composition. The result should feel like a live-action shot where only the foreground subject is paused while the entire world behind them continues naturally.
Paste the full prompt unchanged
Keep the "frozen foreground" block
Keep the locked-camera block
Generate, download & enjoy
Pro Tips
PUSH IT FURTHER
👥
The Duplicate Recipe
ONE HERO, MANY CLONES
Every strong shot follows the same shape: one copy of the character placed front-and-center as the "hero" the camera frames, then 2–3 more copies scattered around doing clearly different actions — driving, biking, cooking, on the phone. The contrast in poses is what sells the duplication. If all the clones do similar things, the trick reads as a mistake instead of a choice.
🪪
Keep The Tag Line
"MAINTAIN THE CHARACTER IMAGE"
Always end the GPT prompt with 4:5 ratio, maintain the character image. That closing instruction is what forces every clone to share the same face, hair, and outfit instead of drifting into different-looking people. Drop it and you'll get five strangers in one frame. The 4:5 ratio also keeps the shot in a clean vertical-friendly crop for social.
🧊
Freeze The Front
DON'T TRIM THE NO-MOVEMENT BLOCK
The whole effect rests on the foreground copy being a perfect statue while the others move. That exhaustive "no blinking, no breathing, no micro-movements" block in the Seedance prompt is doing the heavy lifting — especially the no-blink line. If the front copy flickers even once, the illusion breaks. Regenerate and re-emphasize that the closest character is fully frozen.
🎬
Vary The Motion
DESYNC THE BACKGROUND CLONES
Keep the line "every character should move differently to avoid synchronized animation." If the background copies move in lockstep it looks like a glitchy clone army; staggered, independent motion makes them feel like separate moments of the same person living at once. Run a few passes and pick the take where the timing feels the most natural and alive.