Vidu Q4 AI Video Generator

Vidu Q4 is an AI video model from Shengshu Technology, the company behind Vidu. You give it an image, or a set of reference images and voice clips, and describe what should happen; it returns a finished video with synchronised sound, including speech and sound effects, so there is no separate voiceover recording or lip-sync pass.

The model launched in early October 2026 as a Preview. Shengshu says the Preview comes before the full version and that user feedback will shape the final release, so details can still change before general availability.

One limit matters before you start: Vidu Q4 does not support text-to-video yet. Every clip begins with at least one image. If you only have a text idea, make a first-frame image first, or use an earlier model such as Vidu Q3, which accepts text prompts.

Output ranges from 540p to 4K and clips run 3 to 16 seconds. Drafting at 540p is a cheap way to check motion and framing, then the version you keep renders at 2K or 4K. In image-to-video the aspect ratio of the source image is preserved, so a vertical photo produces a vertical video without a re-crop step.

Reference-to-video combines 1 to 15 images for characters, scenes and style with up to 3 voice references, so a multi-shot scene can stay consistent inside one request instead of several stitched clips. A typical set is a lead character from two or three angles, a second character, a location and a prop.

It is aimed at AI series and short-drama makers, advertisers and brand teams, social creators, and filmmakers doing previs. The workflow is three steps: upload images and voice references, describe the scene, then pick length and resolution and generate.