Image to Video

Image to video keeps the still you already have and adds motion to it. Instead of describing a scene from nothing, you supply the first frame and describe only what should change: the subject turns its head, the camera pushes in, the water starts moving. The model preserves the composition you chose and animates from there.

Start from an image that gives the model something to hold

A sharp, well-lit image with a clear subject animates far more reliably than a busy or dark one. Two practical points:

• Leave room for the motion. If the subject is cropped tight against the edge, a push-in has nowhere to go. A little headroom gives the camera room to move.
• Match the aspect ratio to your target. Cropping after the fact is more work than starting with the frame you intend to publish.

Provide the image by public URL or by uploading it in the form; the file is stored so the render job can read it.

Describe only the change

The prompt should read as a direction to a camera operator, not as a second description of the picture. "Slow dolly in, she turns to look at the lens" is a good prompt for an image-to-video job. Restating the whole scene wastes the prompt on information the model can already see.

If the motion drifts in a direction you do not want — the subject morphing, the background crawling — name the problem in the negative prompt rather than adding more positive instructions.

Settings that change the result

• Model route — the routes offered in the form are the ones reachable at that moment. Live controls are the source of truth, not this page.
• Resolution and duration — both increase the credit cost, which is shown beside the form before you commit.
• Seed — fix it to compare two prompts against the same starting frame; leave it empty for a fresh attempt.
• Negative prompt — the place for artefacts you want suppressed, such as blur or warping.

Where it fits, and where it does not

Image to video suits product shots, portraits, illustrations, and any frame you already like and want to bring to life. It is the wrong tool for a shot that does not exist yet — for that, use text to video — and it cannot invent a different camera angle from the one in your still.

Treat the first render as a test. Quality varies by route and by how much motion you asked for; large changes to a static frame are the hardest case.

Questions people ask

What formats can I supply? Publicly reachable image URLs are simplest; the upload control accepts the common web formats.

How much does it cost? The credit cost appears beside the form before you submit and depends on the model, resolution and duration. Credit prices are on pricing.

Where do I start? Open the Wan workspace and choose the image-to-video tab, or begin with text to image if you still need a first frame.