Text to video AI: write the scene, get the shot

Think of the prompt as one line of a shot list. The model has only a few seconds to tell the story, so give it one subject, one action and one camera move.

A reliable order is subject, action, setting, light, camera and sound. For example: “A red kite dancing over a green hill, a child running below it, late afternoon sun, low-angle tracking shot, wind and laughter.”

• Subject: who or what the shot is about
• Action: what happens during the clip
• Setting: place, weather and time of day
• Light and mood: golden hour, neon, overcast
• Camera: dolly-in, pan, orbit, drone, handheld
• Sound, with Seedance 2.5: ambience, effects or music

Use text to video when you have no footage and want the model to design the whole shot. When the look must match a specific photo, character or product, start from that image instead: image to video keeps the first frame exactly as you uploaded it.