Associate Director, Institute of Artificial Intelligence, Tsinghua University,Biodata TechnologyFounder and Chief Scientist Jun Zhu released on March 29 a highly controllableVideo mockup Vidu Q1, officially claimed to be the industry's firstA large model of a highly controllable AI video.
1AI was officially informed that Vidu Q1 has achieved remarkable results in multi-subject detail controllability (especially motion and layout controllability), sound synchronization controllability, and picture quality enhancement. Taking multi-subject detail controllable as an example, on the basis of semantic commands, by incorporating visual commands of reference diagrams, Vidu Q1 supports the control of the scene'sFurther control of the position, size, trajectory and other attributes of multiple subjects,Precise adjustment of all movement behaviors (exit, exit, sitting and standing postures, courses of action).
It supports sound synchronization and control, which ensures that Vidu Q1 is able to synchronize with the video environment and screen transitions.Output to generate corresponding sound effectsand can accurately control the length of the interval of each sound effect, and accurately set the point in time at which each piece of audio appears.e.g. 0-2s wind sound, 3-5s rain sound, etc..