Alibaba has released Wan 3.0 in beta, doubling its predecessor's maximum generated-video length to 30 seconds and accepting a broader mix of source material. A prompt can combine text with as many as ten images, five videos and five audio clips, or draw from web pages, PDFs and PowerPoint files.
The model can recommend a clip length and extend an existing video. Alibaba says it improves consistency for referenced characters, objects and layouts, areas where generated video often drifts. Those are company claims, and users will need to test difficult motion, faces and interface footage rather than assume every 30-second output remains stable.
Wan 3.0 is available through wan.video, Alibaba Cloud Model Studio and the Qwen Cloud API. Standard pricing per second is listed at $0.05 for 480p, $0.10 for 720p and $0.20 for 1080p during the current discount. A 30-second 1080p video therefore costs $6 on Standard or $8.40 on the faster Prime tier. Availability is beta, so pricing and behavior may change.