Recently, Shengshu Technology officially released a preview version of its next-generation video generation flagship model, Vidu Q4Preview. This update brings significant improvements in generation quality, multi-modal references, and image precision, marking another key step forward for domestic video large models on their path to film-grade quality.
In terms of core features, Vidu Q4Preview fully supports two mainstream modes: image-to-video and reference-based video generation. The image-to-video mode allows high-quality footage to be generated naturally based on the images and text descriptions provided by users; while the reference-based video generation mode supports the creation of videos by combining multiple reference images and audio, greatly enhancing the control options for creators.
This update has made a major breakthrough in multi-modal reference capacity and detail control. The model supports up to 3 segments of reference audio input, allowing the voice tone of the character to remain highly consistent in the video through precise sound references; at the same time, it supports up to 15 reference images, providing a more sufficient visual basis for character facial features, complex scenes, and creative elements such as props, ensuring consistency among elements in long shots and continuous narratives.
In terms of image quality and shot performance, Vidu Q4Preview has further optimized dynamic camera movements, cutaways, and the transition between complex shots. The smoothness and dramatic tension of the footage have been significantly improved. In terms of specifications, it supports output at up to 2K and 4K resolutions, with 10-bit color depth, and can generate videos up to 16 seconds long in a single session. It is perfectly suitable for diverse application scenarios such as AI short films, advertising marketing, social interaction, and professional film production.
In addition, the official platform has introduced flexible specifications and pricing plans tailored to different stages of creation needs. The model supports various resolutions such as 540P, 720P, 1080P, 2K, and 4K, with a launch price starting at 0.09 yuan per second. Creators can begin with lower-cost versions of low-resolution tests, scripts, and footage, and once their ideas are mature, they can directly output high-resolution files in 2K or 4K with fine textures and impressive lighting effects.