On August 24, Alibaba’s video generation foundation model Wan3.0 officially launched. The model has been comprehensively upgraded across generation duration, all-purpose creation, comprehensive reference capabilities, and real-world fidelity. It can generate videos of up to 30 seconds in a single request and, for the first time, supports document inputs in formats including doc, xls, ppt, pdf, and md, striving to accurately recreate the real world. Since the start of its public beta, creators who have witnessed every update to the Wan model family have called Wan3.0 “the most pleasant surprise yet.” Feedback from enterprise users has been remarkably consistent: “stable, realistic, and high-quality” are the three keywords mentioned most often.
Wan3.0 maintains consistency across fine-grained details, including characters, props, sound, spatial relationships, and visual style. This is especially important for enterprise users focused on productivity. For example, in this high-speed car chase, Wan3.0 understands the reference material and handles long-form temporal narratives. Over the 30-second sequence, the characters, vehicles, and environment remain highly consistent. From boarding the car and the ensuing chase to high-speed stunts and the final explosion, the entire sequence is generated coherently—without endless rerolls or repeated rounds of tweaking.
(Video source: LibTV)
Realism is the dividing line that allows Wan3.0 to move beyond the typical “AI look.” Some enterprise users have noted that Wan3.0 fills a long-standing gap in AI video: the ability to consistently produce footage with a convincing human feel. To move beyond the greasy, cookie-cutter appearance of AI-generated people, Wan3.0 strives to give every face its own identity. It captures facial features and skin details more sensitively, while keeping emotional expressions natural and restrained, with micro-expressions and body movements working in sync.
For example, in this character video, four notoriously difficult physical attributes to reproduce—the reflection in the girl’s eyes, hair detail, the movement of the white veil, and the sheen of the water—appear together in a single shot. Wan3.0 maintains a cool gray-blue color palette throughout, while the skin avoids any plastic-like appearance.
(Video source: Flova AI)
A high-quality look means that Wan3.0 has a strong sense of cinematic aesthetics: it focuses on harmonious color combinations while preserving the clarity and fluency of visual storytelling. Users have noted that Wan3.0 can effectively understand and integrate stylized reference images, quickly adapting to the specified visual style.
For example, this stylized short film combines rich red tones with the mechanical aesthetics of steampunk. The weight of the layered skirts, the cold rigidity of the metal gears, and the warm diffuse reflection of candlelight… Different materials retain their individual physical properties within a unified color palette, creating a sense of depth reminiscent of a stage play.
(Video source: 三代科技)
During the public beta, Wan3.0 had already entered enterprise production workflows, supporting large-scale content production in short-form dramas, film and television, and advertising. It has also been integrated into creator platform workflows and used to support cultural and tourism communications and music video production. One user commented that Wan3.0’s core value lies not in the visual impact of individual frames, but in its stable, continuous, and practical content production capabilities. For short-form drama creators, it is “the best value for money.”
Starting today, users can try Wan3.0 on Alibaba Cloud Model Studio, the Qwen AI platform, the Wanxiang official website, and the Qwen app. Alibaba Cloud Model Studio and the Qwen AI platform are offering a limited-time API promotion: from August 24 to September 23, the Wan3.0 Standard Edition is available at 30% off. After the discount, the prices for 480P, 720P, and 1080P are RMB 0.21/second, RMB 0.42/second, and RMB 0.84/second, respectively. Wan3.0 is also now available on Flova, LibTV, 巨日禄, 海艺, and other platforms.