The trick is to change the world in two steps. First you make one still frame of the new world from a frame of your clip, then you hand the video model both the clip and that frame.
Pick a short shot where your face is clearly visible and cut 10 seconds.
Take a still from the clip and turn it into the new location with an image model (Nano Banana 2.1). For a new outfit, attach your character sheet as well.
In Dreamina: AI Video → Seedance 2.5 Flash → Omni reference, 4:3, 720p, same length as the clip. Attach the clip and the reference frame, paste the video prompt.
Cut the two results into one 10-second clip on the original timing: 0–5 s Alpine ridge, 5–10 s Dubai. Export everything at 24 fps so the original and the result stay frame-locked.
In Dreamina an “@” reference chip always jumps to the start of the prompt. Put all the chips first, then refer to them in words: “the first reference is the source video, the second is the new location”.
The original clip
Sultaniye, Turkey · 10 s · filmed on a selfie stick from the car's sunroof
The final clip
0–5 s Alpine ridge · 5–10 s Dubai at sunset
02 / BEFORE AND AFTER
One clip, two worlds
Left of the handle is the original clip, right of it is the final cut. The face, the shout and the laugh never move — only the world around them changes halfway through.
ORIGINALFINAL CUT
Drag the green handle. 0–5 s Alpine ridge · 5–10 s Dubai at sunset.
03 / IN THE CUT · 0–5 S
Alpine ridge
Same shout, same laugh, same camera shake — but the pine road becomes the crest of an Alpine ridge with a sheer drop into the valley on the right.
Step 1 — the reference frame
Reference frame made from a still of the clip · Nano Banana 2.1
Step 1 · Reference frame · Alpine ridge
Prompt text
Model: Nano Banana 2.1 (Magnific), 4:3, 2K.
Image 1 = the first frame of your clip.
Edit Image 1. Keep the man exactly as he is: his face, expression, head angle, navy long-sleeve top, the arm resting on the roof, the white car roof and the open sunroof, the ultra-wide selfie-stick framing and the camera height.
Replace only the surroundings: a narrow asphalt road running along the crest of an Alpine mountain ridge. On the right side of the road the ground ends in a sheer rock drop into a deep green valley hundreds of metres below, with a small village and a river at the bottom. Snow-capped peaks and glaciers on the horizon, a few clouds below the summits, clear late-afternoon sun from the left.
Light on his face and on the car matches the new sun direction. Same wide-lens distortion as the original. Photoreal, no text.
Step 2 — the video
Step 2 · Video · Alpine ridge
Prompt text
Dreamina → AI Video → Dreamina Seedance 2.5 Flash → Omni reference. 4:3, 720P, 10 s.
@Video 1 = your 10-second clip. @Image 1 = the reference frame from step 1.
HARD RULE: @Video 1 is a selfie filmed by the man himself on a stick camera. Never show the camera, the stick or anyone else filming.
Keep the man in @Video 1 exactly: his face, every expression, the shout towards the lens in the middle of the clip, the laugh, every head turn, his navy long-sleeve top, his arm on the roof, the white car and the open sunroof. Keep the camera position, the wide-lens look, every shake and the timing.
Replace only the world around the car with the location from @Image 1: the car drives along the crest of an Alpine ridge; on the right side the road ends in a sheer drop into a deep valley far below, snow peaks on the horizon. The road, the valley and the mountains slide past at the same speed as the road in @Video 1. Wind moves his clothes the same way as in the original.
Sunlight from the left like in @Image 1; his face and the car roof are relit to match.
The same face from the first to the last frame. Photoreal, same grain and sharpness as an action camera.
Keep the original audio: wind and his voice.
[NEGATIVE] pine forest, Turkish road, flat terrain, new cuts, changed facial performance, different face, plastic skin, extra people, visible camera or selfie stick, on-screen text
04 / IN THE CUT · 5–10 S
Dubai at sunset
Three changes at once: the outfit from the character sheet, a black Rolls-Royce instead of the white car, and a downtown Dubai avenue at golden hour.
Step 1 — the reference frame
Reference frame made from a still of the clip · Nano Banana 2.1
Step 1 · Reference frame · Dubai at sunset
Prompt text
Model: Nano Banana 2.1 (Magnific), 4:3, 2K.
Image 1 = the first frame of your clip. Image 2 = your character sheet.
Edit Image 1. Keep the man's face, expression, head angle, the arm resting on the roof and the ultra-wide selfie-stick framing exactly.
Change three things:
1. Outfit: dress him as in Image 2 — royal-blue baseball cap with a white oval patch, round dark tortoiseshell glasses, loose charcoal T-shirt over a white long-sleeve top. Take only the clothes and accessories from Image 2, not its grey background or its panels.
2. Car: he now stands up through the open sunroof of a black Rolls-Royce Cullinan with a glossy black roof; the roof and the sunroof edge are in the same place as the white roof in Image 1.
3. Place: a wide multi-lane avenue in downtown Dubai at sunset, glass skyscrapers on both sides, the tallest tower in the city on the horizon, warm orange sun low behind the buildings, palm trees in the central reservation.
Warm golden rim light on his face and cap from the low sun. Same wide-lens distortion as the original. Photoreal, no text, no logos except the car's own badge.
Step 2 — the video
Attach your character sheet as a third reference — that is what changes the clothes.
Step 2 · Video · Dubai at sunset
Prompt text
Dreamina → AI Video → Dreamina Seedance 2.5 Flash → Omni reference. 4:3, 720P, 10 s.
@Video 1 = your 10-second clip. @Image 1 = the reference frame from step 1. @Image 2 = your character sheet.
HARD RULE: @Video 1 is a selfie filmed by the man himself on a stick camera. Never show the camera, the stick or anyone else filming.
Keep the man's face and performance from @Video 1 exactly: every expression, the shout towards the lens in the middle of the clip, the laugh, every head turn, the arm on the roof. Keep the camera position, the wide-lens look, every shake and the timing.
Change three things in every frame, matching @Image 1:
1. His outfit becomes the one from @Image 2: royal-blue cap with a white oval patch, round tortoiseshell glasses, charcoal T-shirt over a white long-sleeve top. The cap and glasses stay on for all 10 seconds.
2. The white car becomes a black Rolls-Royce Cullinan; he sticks out of its open sunroof.
3. The road becomes a downtown Dubai avenue at sunset: glass skyscrapers sliding past on both sides, the tallest tower on the horizon, palm trees, warm low sun.
The city moves past at the same speed as the road in @Video 1. Warm sunset light on his face and the glossy black roof; reflections of the towers glide across the roof.
The same face from the first to the last frame. Photoreal, same grain and sharpness as an action camera.
Keep the original audio: wind and his voice.
[NEGATIVE] pine forest, white car, navy top without cap, no glasses, new cuts, changed facial performance, different face, plastic skin, extra people, visible camera or selfie stick, on-screen text
05 / WHAT MADE IT WORK
Small rules, big difference
Say it is a selfie on a stick and forbid showing the camera — otherwise the model invents a cameraman.
Name the performance beats you need to keep: the shout, the laugh, every head turn.
Tell the background to slide past at the same speed as the original road.
Relight the face for the new sun direction, or the swap looks pasted on.
Close the leaks in the negative: the old forest, the old car, a different face.