El modelo de generación de videos Tongyi Wan2.7-Video se lanza: admite entrada multimodal completa de texto, imagen, video y audio

Phoenix.com Technology News, on April 3, recently, the video generation model Wan2.7-Video was officially launched. This model supports text, image, video, and audio full-modal inputs, focusing on editing and modification capabilities after video generation, aiming to make video content editable like a document.

In video editing, Wan2.7 supports making local adjustments to the visuals via instructions without needing to regenerate the entire clip. Specific functions include: adding and deleting elements (such as “delete the train in the video”); replacing objects (such as “replace the film with a plate”); modifying object attributes (such as changing the building color); and supporting the precise addition of content based on reference images. At the environment and style level, while keeping the character’s actions unchanged, the background season can be changed from summer to deep autumn, or converted to an art style like felt wool with a single click. In addition, it supports video quality enhancement (such as black-and-white colorization), visual understanding tasks (such as subject segmentation), and adjustments to the shooting method (such as changing focus).

In storyline modification, Wan2.7 allows modifying character dialogue, actions, and even the camera viewpoint without changing the original identity and setting. Dialogue changes can keep emotions, lip shapes, and voice timbre consistent; action changes such as “the girl sitting on the sofa becomes standing and playing games.” On the story level, the player character can be replaced with a medieval knight while maintaining the original gripping posture. Camera angles, shot types, lens types, focal length, and other parameters can all be adjusted via instructions—for example, “change the lens to gradually rise upward from the ground.” These functions are designed to lower the technical barrier for secondary video creation, providing creators with more flexible control methods.

(Editor-in-charge: Dong Pingping)

     [Disclaimer] This article only represents the author’s personal views and is not related to Hexun.com. The Hexun website remains neutral regarding statements, opinions, and judgments made in the text, and does not provide any express or implied guarantee regarding the accuracy, reliability, or completeness of any content included. Please regard this as for reference only, and assume all responsibility yourself. Email: news_center@staff.hexun.com

Report