Before you begin
A 30-second, single-take 9:16 video where the camera chases one object through a whole city and lands back on your opening frame, so it plays as a seamless loop.
Tools used
- Luma (lumalabs.ai)
- Seedance 2.5
- Any image model for the start photo (I used Nano Banana Pro)
Requirements
- A Luma account with access to Seedance 2.5
- One vertical photo of you (or your character) to use as the reference
On this page · 7 steps
Your progress
0 of 7 steps
The walkthrough
Step 01 / 07
Make your start photo
Everything starts from one 9:16 photo: you (or your character) seen from just behind the shoulder, holding a white paper plane, about to throw it toward a city. This frame becomes both the first and the last frame of the video, so make it look exactly how you want the loop to begin and end. Generate it with any image model that takes a reference photo. Upload a clear photo of yourself as the reference and use the prompt below. Keep the face, hair and outfit consistent with your reference.
Exact prompt
Cinematic vertical film still, 9:16, over-the-shoulder from just behind the person in the reference photo (same face, hair and outfit; we see the back of the head and one shoulder at the left edge of frame). They stand on the wet flat roof of a mid-rise building in Tokyo at blue hour, light rain, holding a folded white paper plane up at shoulder height, arm drawn back, about to throw it forward toward the city. Beyond the roof edge: a dense canyon of buildings with soft magenta, cyan and amber neon glow and blurred bokeh, a train line far below, no readable signs or logos anywhere. Shot on ARRI Alexa 35, 24mm lens at f/2.8, shallow depth, rain droplets catching light, wet reflective rooftop, cool teal shadows with warm neon highlights, fine film grain, natural and real, not stylised.
Step 02 / 07