Alibaba has unveiled Wan 3.0, its new AI model for generating videos from written instructions, images, clips, audio, or documents. The tool, already available in public beta through Model Studio, allows users to create scenes of up to 30 seconds with integrated camera movement, dialogue, and sound effects. This launch comes just weeks after the Chinese company introduced Qwen 3.8-Max, its high-end language model, and reinforces its commitment to generative multimedia content.
The most striking feature of Wan 3.0 is its ability to transform a PowerPoint presentation, PDF, or webpage into a narrated video. The user simply writes a prompt, attaches the reference material, and the AI takes care of creating an audiovisual piece with spoken explanations, animations, and automatic staging. This feature opens the door to converting technical documents or training materials into videos without the need for manual editing.
Multiple references and extended duration
One of the most significant improvements in Wan 3.0 over its predecessor is support for multiple reference types. The model can combine up to ten images, five video clips, and five audio files in a single request, along with a document or webpage that provides context. This allows for the creation of complex scenes with much greater narrative coherence than in previous versions.
The maximum clip duration has doubled, from 15 seconds in Wan 2.7 to the current 30 seconds. However, users can manually adjust the length second by second, or let the system recommend an appropriate duration based on the complexity of the instruction. Available output resolutions are 480p, 720p, and 1080p, with fixed or adaptable aspect ratios for different platforms.
Automatically generated and synchronized audio
Another significant new feature is that Wan 3.0 generates audio by default. Characters can speak or sing in sync with the scene, and the AI also incorporates ambient sounds that match the visual content. This capability is especially useful for explainer videos, short advertisements, or training materials, where the coordination between image and sound is essential. If preferred, users can also request a clip without audio and add a voiceover or soundtrack later.
Lip-syncing, movement, and sound effects have been significantly improved, reducing the inconsistencies that were common in previous generations. This makes Wan 3.0 a practical tool for creating video tutorials, software demonstrations, or social media content without needing a production team.
Document conversion and limitations
Wan 3.0's flagship feature is the conversion of presentations and documents into narrated videos. The AI can read a PDF, presentation, or webpage and use its content as the basis for generating an audiovisual sequence. For example, a PowerPoint presentation with several slides can be transformed into a video that explains its main points through cutscenes, on-screen text, and voiceover. However, there is a significant limitation: only one document or webpage can be processed at a time, not both simultaneously.
Alibaba has also improved the rendering of digital elements, such as software interfaces, motion graphics, and in-image text. This type of content is often problematic for video producers, but WAN 3.0 reduces distortion between frames, making it easier to create technical and training videos with acceptable visual accuracy.
Availability and professional orientation
Wan 3.0 is available in public beta through Model Studio, Alibaba Cloud's AI development platform. As it is still in the testing phase, performance and generation times may vary. The company is targeting this tool at professionals who need to produce short social media ads, robot training materials, or convert technical documents into narrated videos. It also envisions its use in the corporate sector, where automatically generating videos from presentations can save hours of work.
The launch of WAN 3.0 comes at a time when AI-powered video generation is gaining prominence. Unlike other models that are still in the experimental phase, Alibaba has opted to offer public access from day one, allowing users to test the tool and report bugs. This strategy contrasts with that of other competitors, who have been more cautious in launching their systems.
In short, Wan 3.0 presents itself as a versatile solution for creating videos from multiple sources, with synchronized audio and a maximum duration of 30 seconds. Its ability to convert documents into audiovisual pieces makes it an attractive option for businesses and content creators, although the limitation of only one source per generation and the maximum resolution of 1080p leave room for improvement. The public beta is already active, and it will be interesting to see how the model evolves in the coming months.