On April 3, 2026, Alibaba's Tongyi Lab officially launched the Wan2.7‑Video video‑generation model series. This follows Wan2.7‑Image and represents another major upgrade in the Wan series. The models support text, image, video and audio multimodal inputs and include four professional models—text‑to‑video (Wan2.7‑t2v), image‑to‑video (Wan2.7‑i2v), reference‑to‑video (Wan2.7‑r2v) and video editing (Wan2.7‑videoedit)—extending AI capabilities from single‑asset generation to the full creative chain, achieving a leap from "generating" to "directing" and being hailed in the industry as a milestone that "makes video as editable as a document."
Users can experience Wan2.7‑Video immediately on the Tongyi Wan official site (tongyi.aliyun.com/wan), the wan.video site and the Alibaba Cloud Bailian platform; the Qianwen app will also be integrated soon. The full model series supports 720P and 1080P outputs, and video length can be specified anywhere from 2 to 15 seconds, covering high‑frequency scenarios such as short videos, short‑drama storyboards, advertising and instructional videos.
Four professional models building the full creative chain
The biggest highlight of Wan2.7‑Video is its full‑chain controllability, which addresses the long‑standing industry pain point that AI video is "easy to generate but hard to edit":
- Text‑to‑video (Wan2.7‑t2v): Creates videos from scratch, supports very long and complex prompts, and delivers natural semantic understanding with coherent storytelling.
- Image‑to‑video (Wan2.7‑i2v): Starts from an image and accurately preserves the main subject, composition and style while generating dynamic video.
- Reference‑to‑video (Wan2.7‑r2v): Accepts multiple reference images/videos as protagonists or objects to produce single‑performer or multi‑person interactions with high character consistency.
- Video editing (Wan2.7‑videoedit): The core killer feature that enables natural‑language instructions to perform local edits on generated videos, such as "remove the train from the footage," "replace the film with a plate," "change the background from summer to late autumn," or "convert to wool‑felt style with one click," without having to regenerate the entire clip.
The model also introduces the instructional "P" video capability: users can precisely control scene structure, local details, temporal changes and character actions via text commands, enabling fine‑grained operations such as adding/removing elements, replacing objects, modifying attributes and adjusting shooting parameters (focus, camera movement). It also supports video enhancement tasks (e.g., colorizing black‑and‑white footage) and subject segmentation, making post‑editing as flexible as image editing software.
Technical breakthroughs: multimodal input and comprehensive style and audio upgrades
Wan2.7‑Video is trained on a native multimodal unified framework, supporting text, image, video and audio inputs to realize a true WYSIWYG creative experience. Visually, it can generate thousands of style combinations covering 2D cartoons, 3D animation, ink painting, clay, cel‑shading and other visual languages; on the audio side, it achieves more natural audio‑visual synchronization and timbre matching to enhance immersion.
Compared with its predecessor, Wan2.7‑Video shows significant improvements in motion amplitude, adherence to physical laws and artistic texture, producing videos with more "director‑level" controllability and professional quality. Tongyi Lab says the series aims to remedy AI video shortcomings in narrative, plot controllability and post‑editing, enabling ordinary users to achieve professional‑grade creation with ease.
Integration and user experience
The model is available via API on Alibaba Cloud Bailian for enterprise integration; individual users can try it on the official website or apps with no barriers. Tongyi Lab has also opened partial developer interfaces to encourage developers to build vertical applications. Pricing for API calls is based on generated duration and resolution, offering clear cost‑effectiveness advantages.
Far‑reaching impact: productivity leap for short‑video and film production
For content creators: Wan2.7‑Video greatly lowers professional thresholds. Independent creators and short‑drama teams can quickly generate customized storyboards and special effects; e‑commerce practitioners can turn a single product image into multi‑scene demonstration videos; educators can easily produce dynamic teaching materials. The "video‑as‑editable‑document" feature reduces iteration costs to near zero and unlocks creative potential.
For the industry ecosystem: In areas such as short video, advertising and film previsualization, Wan2.7‑Video will accelerate the fusion of AI and content production. Traditional shooting and post‑production workflows can be partially replaced by AI, driving low‑cost, high‑efficiency content growth. It also provides strong technical support for game NPCs, virtual anchors and metaverse scenarios, further expanding the boundaries of AI visual applications.
For global AI competition: As a leading domestic multimodal video‑generation model, Wan2.7‑Video forms a distinctive advantage in controllability and full‑chain capability and complements international competitors like Sora and Veo. China's emphasis on pragmatic, high‑utility approaches is again validated, offering developers worldwide more accessible and flexible creative tools.
Challenges and outlook: Despite its leading performance, Wan2.7‑Video as a new model still has room for optimization in generating very long videos and complex multi‑character narratives. In the future, Tongyi Lab may further open more weights, extend video length and deepen integration with Wan2.7‑Image to achieve seamless collaboration between images and video.
The release of Wan2.7‑Video marks a qualitative shift in AI video generation from "being able to perform" to "being able to direct." It not only brings "director‑level" creative capabilities to ordinary users but also signals a fundamental change in content production paradigms—AI is no longer a mere generation tool but an intelligent collaborator that understands creativity. As generative AI continues to penetrate, this model helps China solidify a leading position in the global visual race and injects new momentum of "controllability" and "practicality" into the whole industry.
▶️ Enjoy this article? Creating is hard work!
Follow our YouTube channel:
AI Rosetta
Your likes, subscriptions and shares are our greatest support! 🙏
More AI news at: https://www.airosetta.com