Vidu Q1 is a high-performance video foundation model developed by Shengshu Technology in collaboration with Tsinghua University. Built on the Universal Vision Transformer (U-ViT) architecture, the model integrates diffusion techniques with transformer-based scaling to generate temporally consistent, high-fidelity videos. It supports resolutions up to 1080p and is designed to provide professional-grade cinematic quality for creators.
The model features a versatile multimodal framework that supports text-to-video, image-to-video, and a specialized Reference-to-Video mode. This allows users to provide up to seven reference images to maintain subject and scene consistency throughout a sequence. It also utilizes a "First-to-Last Frame" system to ensure smooth, natural transitions between specific visual anchor points.
One of the defining capabilities of Vidu Q1 is its integrated audio generation, which produces 48 kHz high-definition sound effects and background music tailored to the visual context. The model excels in diverse styles, ranging from realistic live-action to intricate anime, and is optimized for complex motion synthesis and spatial-temporal coherence.