How to Write Image-to-Video Prompts

Image-to-video works best when you treat the source image as a locked first frame and describe motion progression, not a brand-new scene. Start with identity and environment anchors, then add camera intent and emotional pacing.

Formula: [source frame subject] + [environment motion] + [camera path] + [continuity anchors] + [quality constraints]. Example: “From this noble girl portrait at a side gate, animate lantern flicker and silk sleeve motion, then slow dolly-in to token close-up while preserving eye color and brocade details.”

1) Define the still image subject and setting. 2) Describe first movement in 1-2 actions. 3) Specify one camera movement only. 4) Add continuity anchors (face, hair, outfit, prop). 5) Add negative prompt and render 3-6 seconds first.

Ancient Palace Night Escape from still frame, black horse under lantern light, noble girl at side gate, gentle push-in, stable facial identity.

Fantasy Forest Reunion from illustration, drifting leaves, over-the-shoulder reveal, pendant token close-up.

Lonely Boy Singing at Desk image-to-video, notebook page flutter, soft profile-to-close-up transition.

Mistake 1: asking for extreme camera movement from a static image in one pass. Mistake 2: forgetting to restate identity anchors after movement cues. Mistake 3: adding unrelated props not present in source frame.

Sora: use explicit temporal sequence words. Vidu: keep subject/mood compact and clear. Kling: add physical motion verbs. Runway: include shot labels and transition intent. Veo: describe atmosphere, materials, and weather for richer context.