At the World Artificial Intelligence Conference (WAIC) 2024.SenseTimeControlled characters launchedVideo Generation of Large Models——VimiThe Vimi model utilizes Shangtang's advanced large modeling technology to be able to generate character videos that are consistent with the target's movements from a single photo, enabling precise expression and body control. The model supports a variety of driving methods, including video, animation, sound, and text, and features facial tracking technology and precise control of details accumulated over many years, enabling it to generate video content with high consistency and harmony of light and shadow.
Vimi is particularly stable, capable of generating single-camera character videos that grow to over one minute without deteriorating over time. It can also adjust the environment according to the character's movements, simulate changes in camera angle and hair shake, and provide realistic visual effects. In addition, Vimi also supports light and shadow change simulation, providing video creators with a wealth of creative freedom.
Vimi CameraAs the first C-end application based on Vimi's large model, it is mainly aimed at female users to meet their entertainment creation needs. After uploading high-definition pictures of people from different angles, Vimi Camera can automatically generate digital doppelgangers and different styles of portrait videos, providing a variety of generation styles.Vimi Camera also supports the generation of interesting character emoticons from a single picture, providing a variety of ways of playing and realizing personalized creations.
Currently, Vimi Camera is openingClosed betaInterested users can apply for the experience by following the public number and filling in the reservation link.