Select a clean starting image
Choose an image with the intended composition and no visible defects in faces, hands, or important objects. Its aspect ratio guides image-to-video output.
Choose an image whose composition, character design, and lighting already fit the scene. Image-to-video uses this single opening frame; a separate ending frame is not supported in this mode. Keep the shot description focused on motion, camera position, and emotion. If you need several image references or audio references, choose reference-to-video instead.
Choose an image with the intended composition and no visible defects in faces, hands, or important objects. Its aspect ratio guides image-to-video output.
Open your project and select Vidu Q4 Preview with a single first-frame image. Use reference-to-video if the shot needs multiple references.
Describe the motion and camera move without rewriting the character appearance. Select 3–16 seconds, a supported resolution, and the audio option. Review the current credit estimate before generating.
Check that the first frame is respected, the action reads clearly, and the ending gives you a useful cut point. If something drifts, simplify the movement or improve the starting image before retrying.
Vidu Q4 Preview generated sample
Use the selected image as the opening frame for @Mira at @Station. She slowly raises her gaze toward something just outside the frame, then settles into a still, attentive pose. Keep a medium close-up and a locked camera so the change in expression remains easy to read. Maintain the direction and softness of the existing light. Add only a slight movement in the background and avoid a cut, a new character, or a second action.
Begin from the selected image of @Pendant at @Workshop. A hand enters from the right, gently rotates the object a small amount, and then holds it still. Use a restrained push-in rather than orbiting the object. Preserve the visible proportions, surface markings, and contact with the table. Keep the original lighting direction and let the highlight move naturally across the surface. Finish with a clear, quiet hold that can be used as an editing point.
Start from the selected image of @Garden. Move the camera forward very slowly along the existing path, keeping the horizon level and the main doorway in the same part of the frame. Allow a few leaves to shift in a light breeze while the architecture remains steady. Preserve the calm evening mood and the original light direction. Keep this as one uninterrupted establishing shot, without introducing people, changing the weather, or revealing a new location.
Keep visual identity in the image and in the relevant asset. Use the description for new information: one action, a camera instruction, emotion, and lighting continuity.
Generation can change hands, faces, or object geometry even with a clear image. Compare the opening, middle, and ending, then revise one instruction if the result misses the intent.
Use reference-to-video for 1–15 images and up to 3 audio references. The image-to-video workflow on this page starts from one image. Generated audio can be switched on or off.
Use the same input, prompt, and settings as a starting point. Generation varies between attempts; compare the action, framing, and ending rather than expecting identical frames.
Bring one image, describe one action, and review the generated clip before editing.