made in ChinaVincennes Video Large Modelcompany Vidu Vidu 1.5, a new version of Vidu that achieves a "world-leading" breakthrough in understanding diverse inputs and breaking the "consistency" barrier, was announced yesterday. Vidu 1.5 is now available.
Vidu says that it has been available since its launchRole consistency generation capabilityVidu's newest update, Vidu, solves a key pain point in video generation by targeting people's facial features to ensure consistency. In this update, Vidu is showing for the first timecontextual learningThe ability to enhance the controllability of the overall video model, and the consistent generation of multiple perspectives, subjects, and elements through flexible multiple inputs.
I was told that it visualizes "technological breakthroughs" in three areas:
- Precise control of complex subjects:Whether it's a detailed character or a complex object, Vidu ensures that itsCoherence in many different perspectivesFor example, a complex 3D animation style character can remain in the same position regardless of changes in perspective. For example, a complex 3D animation style character can alwaysEnsure consistent full-angle detailing of head shape and clothing, avoiding the flaws of traditional video models when switching between complex viewpoints.
- Natural consistency of the character's facial features and dynamic expressions:In close-up shots of people, it is able to maintain the consistency of their facial features and expressions in a natural and smooth manner.Avoids facial stiffness or distortion.
- Multi-subject consistency:Users can upload multiple subject images, including character roles, prop objects, and environment backgrounds, and realize the interaction of these elements in video generation. For example, users can upload images of subjects, objects, and environments to create a scene in which a customized character wears a specific costume and moves freely in a customized space.