Vidu

Vidu model adopts the team's original Diffusion and Transformer fusion architecture U-VT, supporting one-click generation of HD video content up to 16 seconds in length, with a high resolution of 1080 P. Vidu opens up the two core functions of text-generated video and graph-generated video, providing two time lengths of 4s and 8s, with a resolution of up to 1080 P. In terms of effects, Vidu adds the ability to generate a 4-second video with a high degree of fidelity, character consistency, and special effects. Vidu in the continuation of high dynamics, high fidelity, high consistency and so on, on the basis of the new character consistency, anime style, text and special effects screen generation and other capabilities.Vidu measured reasoning speed for the generation of a 4-second clip takes 30 seconds.