Ground the Shot
Skip the landmarks. Wet flagstones, weathered brick and polished marble carry more regional truth than any single building.
Create hyper-real 6-second period clips. Drop them straight into your timeline - zero stock footage required.
Generate a China Scene









Prompt Guide
Model
Reference Image
Four layers that make a Chinese street read as a place instead of a backdrop.
A water town lane, a night market, a ridge watchtower, a courtyard. One location per clip. Prompts that ask for the Great Wall and Shanghai in the same shot produce neither.
Wet flagstones, black tile, white plaster, red paper, brass fittings, steam. Materials are what make a scene read as China rather than as a generic old town in another country.
Han or Tang dress, Republican era streets, present day. Rain, snow, morning fog or summer haze. Era and weather together decide whether the clip can sit in a travel edit or a period story.
Ten clips from one AI China video generator: a rain lane, a lantern street, a dynasty courtyard, a modern skyline.
A fortified stone watchtower on a steep ridge before sunrise, low cloud drifting through the valley, crenellations lit orange on one side, slow drone push in, deep blue shadows, epic wide framing.
Try This StyleHundreds of floating paper lanterns drift on a dark river during a festival, warm light on the water, silhouettes of families on the bank, camera slow pan right at water level, night, rich red and gold.
Try This StyleTwo men in long robes argue quietly across a wooden table in an old tea house, porcelain cups, steam, a paper screen behind them, candle and window light mixed, medium two shot, locked camera, restrained period feel.
Try This StyleCommuters stream out of a glass metro entrance in a dense modern city at dusk, neon signage above, wet pavement, scooters passing, camera static at street level, cool blue with warm shop light, cinematic framing.
Try This StyleAn arched stone canal bridge at blue hour, lanterns just lit, still water doubling the arch, slow push forward, 16:9.
Try This PromptA grey brick courtyard gate under falling snow, red door frame, one warm window behind, slow tilt up, muted palette.
Try This PromptClose shot of a dragon dance head weaving past, embroidered silk scales, warm firelight, handheld sway, shallow depth.
Try This PromptDozens of paper sky lanterns lifting over a dark rooftop, low angle looking up, embers fading, slow climb, deep blue night.
Try This PromptA brush tip touching silk paper, black ink spreading in one stroke, candle at the edge, macro shot, no face in frame.
Try This PromptA raised pedestrian walkway in rain at night, headlights streaking below, wet railing catching colour, slow pan right.
Try This PromptThree visual cues that instantly separate a real street from a studio set.
Skip the landmarks. Wet flagstones, weathered brick and polished marble carry more regional truth than any single building.
Warm core, cool edge. Drop a 2700K lantern pool off-centre against a blue dusk, and the contrast reads as cinema instead of stock.
Let bokeh do the work. AI text fails up close, so push signage deep into the background blur, keep the focus sharp and add captions in post.
Notes from editors who shot these streets.
I needed a rain lane for the opening of a travel piece and had no footage from that trip. Wet stone and the lantern reflection did the work. I added the city name as a caption instead of asking for signage.
The wall shot took one reroll to get the mist right. I used it behind a title card for a history video, and the slow ascent left space at the top exactly where I needed it.
My food channel needed a market street behind a voice-over. Steam and hanging bulbs were there on the first try. I kept the lower third clear so my subtitles did not cover anyone's face.
Yes, at short length. You get a recognisable street, gate or skyline in six seconds. Famous landmarks come out approximate and change between runs, so treat them as an impression rather than a document.
Text rendering is the weakest part of every video model. Keep signage small, angled or out of frame, then add your own caption. Short Latin words sometimes survive; long character strings almost never do.
No. This page works from a written description. Upload a photo only when you want a specific building or a person you already have rights to use.
Grok Imagine Video 1.5 runs one free trial without a subscription. The other models on the picker are locked until you upgrade, and each render is short by design.
Check the license of the model that made it before you publish. Avoid real shop names, logos and identifiable faces you do not own, and keep a note of the prompt with the file.