Chinese tech company Alibaba has released a new artificial intelligence model called Wan2.7-Video. The company says the tool is meant to turn everyday creators into directors by handling most of the heavy lifting in video production.
The launch happened only a few days after Alibaba put out its latest picture generation tool, Wan2.7-Image. Releasing both tools so close together shows that Alibaba is moving fast to build a full set of AI products for images and videos.
Wan2.7-Video comes with four different parts. One turns written descriptions into video. Another turns existing pictures into moving footage. The third one uses reference images to guide the output. The last one focuses on editing clips that users already have.
All these parts work together so that someone can write a script, pick a style, and get a finished short video without jumping between different apps. The system can handle clips as short as 2 seconds or as long as 15 seconds. Users can choose either 720p or 1080p quality. For businesses, Alibaba offers special tools to process many videos at once.
One of the main selling points is that creators can manipulate their videos just by typing ordinary sentences. For example, a user can ask the AI to alter a character’s outfit, swap a rainy background for a sunny one, or shift the camera angle from a close up to a wide shot.
The Alibaba Wan2.7 model also keeps lighting consistent across scenes, which has been a common problem with older AI video tools. If a user changes a character’s dialogue, the AI will adjust the lip movements automatically. It can also hold onto a character’s unique voice even when the script is rewritten.
The tool keeps track of up to five different characters in a single story. Each one can have its own look and voice. The AI also supports more than 50 emotional expressions, from happy to sad to angry.
According to Alibaba, users can type a single sentence and get a full storyboard with professional looking transitions in one shot. The AI can add drone shots that feel like a first person view, 360 degree orbits around a subject, and smooth scene changes that don’t feel jarring.
Wan2.7-Video also uses a special method to avoid the sudden stopping problem that many video AIs have. Users can tell the tool exactly what the last frame of a video should look like, and the AI will build toward that ending smoothly.
Alibaba also released a stronger version called Wan2.7-Image-Pro. That one offers sharper understanding of user instructions and can produce 4K images.
Both the video and image tools are available now on Alibaba Cloud’s Model Studio platform and on the official Wan website. They are also built into the Qwen App.