Vidu: Vidu Q3 Turbo Start-End to Video API

Vidu Q3 Turbo Start-End to Video is a Turbo-accelerated start-end frame-driven video generation model of the Q3 series developed by Shengshu Technology. This model adopts a unique start-end frame input paradigm, where users simultaneously provide the starting frame image and ending frame image of a video, and the model automatically infers and generates smooth, reasonable transition video content between the two frames. Under the Q3 Turbo accelerated inference framework, this model prioritizes maximizing generation speed, providing an efficient start-end frame video generation solution for users and platforms highly sensitive to generation timeliness and call costs. Key capabilities include: Dual Anchor-Point Driven Generation via Start-End Frames: Uses the user-provided starting frame and ending frame as dual visual anchor points, automatically inferring and generating transition videos with temporally logical coherence between the two frames, ensuring the generated video visually transitions naturally from the starting state to the ending state, giving users precise control over the content of video beginnings and endings. Turbo Extreme Speed Inference Priority: As the Turbo-accelerated version of the Q3 series, it is designed with maximizing inference speed as its core objective, delivering the fastest start-end frame video generation speed within the Q3 series, significantly reducing user wait times, particularly suitable for high-throughput batch processing scenarios and latency-sensitive real-time content generation needs. Intelligent Transition Content Inference: Based on comprehensive understanding of the semantic content, spatial relationships, and visual style of the starting and ending frames, intelligently infers the most reasonable motion trajectory, scene changes, and content evolution path between the two frames in Turbo acceleration mode, generating visually logically self-consistent transition video content. Text-Assisted Guidance Support: Supports optional text prompt input, using text descriptions to provide additional guidance on transition content, motion style, and scene atmosphere between the start and end frames, providing a degree of generation content controllability under the speed-priority premise. Cost-Efficient Resource Consumption: The Turbo version is the most economical in inference resource consumption, suitable for large-scale concurrent calls, high-frequency batch processing, and commercial application scenarios with limited cost budgets, effectively reducing the per-generation cost of start-end frame video production. Typical use cases include rapid video transition content generation, batch start-end frame animation processing, storyboard dynamic pre-visualization, low-cost video content transition production, rapid creative video prototype validation, and content creation workflows requiring precise control over video start and end frames.