AI video prompt builder.
One idea, two model-ready prompts.

Original prompt · Lantern on a Canal

This is the prompt used for the measured clip above. Results can vary between runs.

integrated_multimodal_description: [Shot 1] Live-action, cinematic, shallow depth of field. A single paper lantern drifts down a narrow canal at night in a southern Chinese water town, seen from a low camera just above the waterline. The lantern's warm orange glow is the only light source near the camera, and it throws a soft moving pool of colour across the wet stone walls on either side. The camera holds a static shot, letting the lantern travel from the upper third of the frame down toward the lower right, then the camera pushes in with small amplitude at slow speed as the lantern passes and its reflection breaks apart on the ripples and reforms. Background: whitewashed houses with dark tiled roofs, a few windows still lit. Light rain has just stopped; the air is thick and the far end of the canal dissolves into mist.

overall_soundscape: Quiet night in a water town. Water laps gently against stone, the rain that has just stopped drips from the eaves into the canal at irregular intervals, and the lantern's paper and thin bamboo frame creak faintly as it turns. A distant dog barks once, and a single bicycle bell rings far away.

non_diegetic_music: A sparse solo guzheng melody at a slow tempo, played with long pauses between phrases, the strings allowed to ring and decay. No percussion.