Create a video from a written scene. Choose a model, describe the action, and use the output settings available for that model.
Describe a subject, action, and setting to generate a video without uploading an image.
Select the resolution, duration, and aspect ratio offered by your chosen model. Available options differ across the Wan models.
Refine one action or camera instruction at a time and compare the generated clips.
Write a prompt describing the subject, action, scene, and camera movement. No source image is required.
Select a workflow from the model menu. Available inputs and settings change with the selected model.
Write a prompt describing the subject, action, scene, and camera movement. No source image is required.
Choose from the resolution, aspect ratio, and duration options shown for your model, then check the credit estimate.
Sign in, submit the generation, and review the result in the preview or generation history. Download the result when it is ready.
Describe what should appear, what should happen, and how it should look. Start with one clear idea.
Subject + Action + Setting + Visual style
Identify the person, object, or scene that matters most.
Describe the movement or change you want to see.
Add the location, lighting, composition, and visual style.
Example prompts are shown in English. You can adapt them to your scene.
A red bicycle stands beside a cafe on a rainy street. The camera slowly moves closer as reflections ripple in the puddles. Soft evening light.
A paper boat drifts down a stream. Close-up, gentle camera tracking, warm morning light.
Text-to-video starts from a prompt. To use an existing image as the starting frame, choose an image-to-video model instead.
Use the options shown after selecting a model. Resolution, aspect ratio, duration, and required inputs vary by workflow; a subscription does not unlock settings the model does not support.
The estimate depends on the selected model and settings. Some reference workflows also account for uploaded video duration. Review the displayed total before confirming generation.
Processing time varies with the model, task size, and service load. Follow the task status in the workspace or open generation history to check the result later.
Keep your source material, adjust the prompt or settings, and submit another generation. Each new generation has its own credit estimate.
