Vidu.

The reference sets the boundaries of the scene

Vidu generates video from text, images and references. These provide different ways to control a scene: describe it from scratch, animate a selected frame or use a character’s distinctive features. The model and selected tool determine which combinations are available.

The official platform also covers images, sound and digital characters. Do not treat all these features as one universal function: check duration, input material and output format for the particular mode. A separate API platform provides programmatic access.

For a first attempt, use a character with a clear silhouette and a simple action. Specify clothing, one important detail and limits on changes. Many characters, complex objects and sudden camera turns in the same scene make likeness preservation much harder to assess.

Example: a short scene with a mascot

Prepare your own mascot illustration and request one action, such as turning towards the viewer or pointing. Keep the camera steady. Check silhouette, eyes and brand elements throughout the clip. Even a small logo error needs fixing, however convincing the movement.

Keep the reference and visual conditions for the next scene. Do not expect independent generations to match lighting and scale automatically. Editing can soften the transition, while text and precise symbols can be added in a separate layer. This keeps the character’s recognisability under the author’s control.

Working conditions

The user-facing service has plans and generation limits. The API has separate billing and task management. Check reference rights, commercial use and model availability before starting a series. A realistic estimate allows for several variations and manual selection, not just one prompt.

Suitable tasks

• Animation of a reference character
• Video from an input image
• Short advertising scenes
• Programmatic video generation