Technology Desk: Alibaba has launched its new model Wan 3.0 in the field of AI video generation. The company says that this model can understand many types of files apart from text and pictures and create videos based on them.
Video up to 30 seconds will be ready
The specialty of Wan 3.0 is that it can generate video up to 30 seconds at a time. In earlier Wan 2.7 models, this limit was 15 seconds. That means the maximum length of the video has been doubled in the new model.
According to the company, the ability to create longer videos will allow users to show an entire story or more information in a single video. This will reduce the need to add separate small clips.
PDF, PPT and Excel will also become references
The most special feature of WAN 3.0 is its multimodal input support. In this, users will not only be able to create videos by writing text, but will also be able to use many types of files as reference.
These include:
pdf
PowerPoint (PPT)
Word document
Excel and other spreadsheets
Image
audio
current video
webpage
For example, by inputting a PPT or PDF of a product, a promotional video can be created based on it. This can make it easier to use AI videos in marketing, product demos and other business tasks.
Will generate audio along with video
WAN 3.0 is not limited to just creating video visuals. Support for audio-visual generation is also provided in it. The model can also create dialogue, background sounds and other audio elements along with the video. Due to this, the user may have to reduce the use of separate tools for video and sound.
Focus on keeping characters and objects more consistent
While making videos with AI, a problem often arises that the character's face, clothes or the design of a product changes in different frames. To reduce this problem in Wan 3.0, special attention has been paid to character and object consistency. The purpose of this is to keep the identity of objects in context and their details more consistent throughout the video.
Video output up to 1080p
According to Alibaba Cloud Model Studio, Wan 3.0 can create videos in 480p, 720p and 1080p resolutions. The starting price of the API varies depending on the resolution. Prices start at around $0.05 per second for 480p, $0.10 per second for 720p, and $0.20 per second for 1080p.
Approval required for API access
Currently, API access of Wan 3.0 is being given after approval. Alibaba Cloud is currently offering it as a public preview. This means that the model's features are available, but the API is not immediately open to every user. Amidst the increasing competition of AI video generation, through Wan 3.0, Alibaba has emphasized on increasing the length of videos as well as creating videos from different files and bringing audio-visual generation into a single system.
