wtf is vidu-s1?

Vidu S1 is a research project that generates video in real time from spoken voice commands. Instead of typing a prompt and waiting for a video to render, a person can speak to a digital character on screen and have it react and perform actions as they talk, with the video appearing live rather than after a delay. The project describes three main capabilities. First, spoken instructions can directly control what a digital character does on screen while the interaction is happening. Second, the system can keep generating video continuously rather than producing a short fixed clip, and it reaches up to 42 frames per second at 540p resolution, a video quality setting, while running on consumer grade graphics cards rather than requiring specialized data center hardware. Third, users can customize the appearance of the character, including realistic people, anime style figures, pets, or other avatar types, along with their voice. This repository itself is mostly documentation and links rather than runnable source code. It points to a hosted website where anyone can try the interactive video feature directly, an academic paper describing the technical approach, and user guides and API documentation in both English and Chinese for developers who want to integrate Vidu S1 into their own applications. There is also an installable skill for AI coding agents like Claude Code that helps automate API integration. Because the core model and heavy computation happen on the company's own servers rather than in this repository, using Vidu S1 in practice means either trying the web platform or calling their hosted API, not running the model locally from this code.