Alibaba Wan 3.0 launches 30-second video generation at 0.6 yuan

Alibaba Group's Wan 3.0 video generation model entered public beta on Aug. 6, producing single 30-second clips at 0.6 yuan per second for 720p resolution, intensifying competition in China's AI video market.

The model "strives to accurately restore the real world — thousands of faces for thousands of people, every frame both real and credible," Alibaba said in its launch announcement, emphasizing visual fidelity rather than raw generation length.

Wan 3.0 supports document inputs including doc, xls, ppt, pdf, and md formats — a first for the model family — alongside text, image, audio, and video modalities. The 30-second single-generation capability lets creators express complete narrative intent in one pass, a step up from the 5-10 second clips typical of earlier video models. At 0.6 yuan per second for 720p, a full 30-second clip costs 18 yuan.

The launch puts Alibaba's video generation ambitions directly against OpenAI's Sora, Google's Veo, and domestic rivals including ByteDance's Jimeng and Kuaishou's Kling. For Alibaba, the model strengthens its cloud AI portfolio at a time when hyperscalers are pouring nearly $600 billion into AI capex globally, according to industry data.

Document Inputs Set Wan Apart

The document-format support is a notable differentiator. Most video generation models accept text prompts, reference images, or audio tracks, but Wan 3.0's ability to ingest structured documents means users can feed it presentation decks, spreadsheets, or PDFs to generate video content directly from business materials. This could appeal to enterprise users in marketing, education, and corporate communications — segments Alibaba Cloud has been targeting as it expands its AI services.

The model's real-world rendering claims also address a persistent weakness in AI video: maintaining consistency across faces, objects, and environments. Alibaba says Wan 3.0 delivers "thousands of faces for thousands of people," meaning each character maintains a distinct visual identity across frames. This matters for commercial applications where brand consistency and character recognition are critical — a spokesperson who looks different in every frame undermines the credibility of AI-generated corporate content.

The 30-second generation window is itself significant. Most consumer-facing video models produce clips of 5-10 seconds, requiring users to stitch multiple generations together for longer sequences. Wan 3.0's single-pass 30-second output reduces that complexity, potentially making the model more attractive for short-form advertising, product demos, and social media content where longer continuous shots are valuable.

Pricing Pressure on Global Rivals

At 0.6 yuan per second for 720p, Wan 3.0's pricing undercuts Western competitors. OpenAI's Sora and Google's Veo have not published per-second pricing, but enterprise AI video tools typically charge significantly more for comparable output. The price point reflects the broader trend of Chinese AI vendors competing on cost — Alibaba's Qwen large language models have similarly priced below Western equivalents, and ByteDance's Doubao models have pushed prices even lower.

For Alibaba, the video model is part of a broader AI monetization push. The company's cloud division has been the primary growth engine, and AI-related revenue has been a key driver. Wan 3.0 adds another product to Alibaba Cloud's AI portfolio, potentially attracting media, advertising, and content creation workloads. The company's stock has benefited from AI enthusiasm, with investors pricing in continued cloud growth as Alibaba expands its AI infrastructure footprint.

The competitive stakes extend beyond video generation. Alibaba's AI investments span large language models (Qwen), cloud infrastructure, and now video generation, pitting the company against both Western hyperscalers and domestic rivals. Nvidia's GPU supply constraints and the broader AI infrastructure buildout remain critical factors in how quickly Alibaba can scale Wan 3.0's deployment. The company's ability to convert AI model leadership into cloud revenue will determine whether the current valuation premium holds.

This article is for informational purposes only and does not constitute investment advice.