wan 3.0: a practical guide to text and image videos

wan 3.0 turns a written scene or a starting image into a video. On Yix, you can choose a duration from 2 to 30 seconds and an output resolution of 480p, 720p, or 1080p. The workflow produces silent clips, so plan to add narration or music during editing if your project needs them.

Open the wan 3.0 video generator to select the model directly. This guide walks through a first shot, explains what changes when you upload an image, and gives you prompts to adapt. The cover is an AI-generated tutorial illustration, not a sample video produced by this model.

What you can control in wan 3.0

Alibaba’s model overview describes text and image inputs, resolutions up to 1080p, and clips up to 30 seconds. This guide covers the text and single-image workflow available in Yix. Settings were checked on September 10, 2026.

In wan 3.0, a longer duration gives an action more time. It does not automatically make a better shot. For a coffee cup by a window, five seconds can be enough for steam and a slow camera move. A 30-second request is more useful when you have an action that needs that time.

Make your first text-to-video shot

Start with one subject in one setting. In the generator, select wan 3.0, leave the image field empty, and describe the visible action. Choose 16:9, five seconds, and 480p for a small first test. Check the credit quote before pressing Generate.

Try this original prompt as a starting point:

One ivory ceramic coffee cup on a wooden cafe table beside a rain-streaked window. Thin steam rises from the coffee. The cup stays still. The camera slowly pushes closer while keeping the whole cup in frame. Soft amber light inside, cool daylight outside. One continuous shot, no cuts, no added objects or lettering.

The cup stays still while the camera moves. That distinction is useful: “make the cup cinematic” leaves both actions open. If the first result moves too much, replace the camera sentence with “The camera remains locked in position.” Keep everything else unchanged so the next comparison tells you something.

These prompts are suggested exercises, not a claim of benchmark results. Generations vary, and a request to preserve an object is not a guarantee that every frame will match it.

Animate a starting image with wan 3.0

Upload a clear JPG, PNG, or WebP image under 10 MB. Choose a source that already has the framing you want. A vertical photograph will guide a vertical shot; selecting a landscape ratio cannot override the image's proportions. Yix shows the image-ratio setting while that reference is present.

For a wan 3.0 shot using an existing photograph of a cup, spend the prompt on motion rather than describing a different cafe:

Animate this starting image as one continuous shot. Keep the cup and table stationary. A thin trail of steam rises gently. Very slow camera push-in, keeping the entire cup visible throughout. Preserve the lighting and background arrangement. No cuts, no hands entering, no new objects.

Image input in wan 3.0 establishes a starting point. It does not lock every later detail. Watch the handle, rim, and any printed design as the camera moves. If those details matter to an advertisement, compare the video with the original photograph before publishing it.

For product work, prepare and review a still in Product Studio, then bring the saved image into the video generator. The product photography prompt guide explains how to keep a label visible when building the source scene.

Choose resolution and credits deliberately

wan 3.0 uses Yix credits according to duration and resolution. The following quotes use the current Yix credit policy for one five-second clip; text and starting-image generation use the same rate.

A two-second 480p clip costs 80 credits. The model configuration defaults to five seconds at 1080p, while switching from another model may retain compatible settings. Read the quote in the generator instead of assuming which setting is selected. Yix pricing lists the available credit packages.

Use a low-resolution attempt to decide whether the composition and motion are worth pursuing. Generating again at 1080p creates another video; it does not simply sharpen the exact frames you approved. Budget for that second attempt and inspect it again. Longer clips also increase the quote.

Give each shot one job

A useful wan 3.0 prompt names the subject, what it does, and what the camera does. Add lighting when it affects the shot. Avoid asking a short clip to introduce a room, follow a person, reveal a product, and end on a close-up all at once.

For a landscape exercise, try:

A small wooden rowing boat tied to a dock on a misty lake at dawn. The boat rocks gently in place. Fixed wide camera, no zoom or pan. Soft grey light, faint ripples across the water. One continuous shot with no scene change.

Here the camera stays still and the boat moves. If the motion looks wrong, shorten the action description or reduce the requested duration. Raising resolution is unlikely to solve a poorly specified action. For a sequence with several distinct beats, create separate shots and assemble them in an editor.

Review the whole clip before downloading

After a wan 3.0 generation finishes, play it through to the end. Look for changing object shapes, unexpected cuts, unstable backgrounds, or a subject drifting out of frame. Pause at the beginning, middle, and end if a small detail matters. An attractive opening frame can hide a problem later.

Use the download button when the result is suitable. Keep the source image and the prompt alongside the downloaded video so you can identify what needs changing on a later attempt. If you need synchronized dialogue, the Seedance guide covers a different workflow available on Yix.

For your next wan 3.0 shot, choose the smallest action that communicates your idea. Start in the generator, set the frame and duration, and check the quote before committing credits. Review the movement before making the clip longer or increasing its resolution.