Wan 2.7 Image-to-Video: First Frames, Timeline Prompts, and Motion

What Wan 2.7 does

Wan 2.7 turns a single still image into a short video — 5 to 15 seconds — and can generate matching audio. What sets it apart from a one-line "animate this" tool is frame-by-frame control: you write a timeline that tells the model what happens at each second, so a clip follows a plan instead of drifting. That makes it a good fit for product demos, short spokesperson clips, and any shot where the motion needs to hit specific beats.

Start with a strong first frame

The first frame defines the starting state — pose, scene, lighting, and overall style. Give the model a good one and everything downstream is easier. Keep the main subject at roughly 60–70% of the frame and leave room around it for motion. Avoid cropped hands, blocked faces, cluttered backgrounds, and unnatural joint angles. Use at least a 1024×1024 input, and skip harsh HDR or heavy sharpening — over-processed frames tend to shimmer once they move.

Write an "At 00:XX" timeline prompt

Wan 2.7 reads an 'At 00:XX' timeline, one line per second. The first line, At 00:00, anchors the shot: shot type, camera angle, subject, pose, scene, lighting, mood, and quality tags. Every line after that describes only what changes — a small motion, then a larger one — rather than repeating the static details. Keep the whole prompt around 50–200 words, and under 5,000 characters.

A clean template, with a product turntable as an example: 'At 00:00, a medium shot at eye level, showing a white sneaker on a matte pedestal, soft studio light, clean and premium, sharp focus. At 00:01, the pedestal begins to rotate slowly. At 00:02, the sneaker turns to reveal its side profile. At 00:03, it completes a gentle half-turn. At 00:04, it settles facing the camera. Camera remains static. No distortion.'

Motion, camera, and audio

Three rules keep motion believable. Use one camera movement per shot — a fully static camera is the safest default. Describe motion with soft speed words like 'slowly', 'gently', and 'gradually', which the model handles more stably than abrupt action. And set the lighting in the 00:00 line so it stays consistent as the shot moves. Wan 2.7 can also generate audio; short, emotion-labeled lines work better than long dialogue.

Built-in prompt optimization

Wan 2.7 has built-in prompt optimization: submit a short prompt and let the model expand it into a fuller one. It is handy for fast drafting. The catch is that expansion can drift on the details you care about, so once you have a direction, lock the important things by hand — subject identity, shot type, pose, camera behavior, and the final timing — rather than leaving them to the auto-expansion.

Known limitations — and how to work around them

Three weak spots are worth planning around. Texture flickering: fine patterns like lace or plaid can shimmer in motion — prefer simpler textures in the first frame and keep lighting stable. Long-video quality drop: very long clips lose detail, so split longer actions into 5–8 second segments, each focused on one motion idea. Facial consistency: vigorous motion can shift a face — add 'face stable, no deformation', reduce the motion amplitude, and keep the camera still. Hands remain a common weak point too; keep gestures simple, keep hands away from the frame edge, and add 'fingers maintain correct anatomy' when hands matter.

FAQ

Related reading