Wan (Alibaba)

Alibaba's open-source AI video model generating cinematic clips from text and images

Hand-reviewed by the Toolglade team·Updated Aug 2026

video#ai-video#text-to-video#open-source#image-to-video

Free plan Free trial Claimed API Self-hosted

Toolglade’s take

Wan is notable for combining frontier-grade AI video with an open-source release, which is rare among top video models. Self-hosting and per-second cloud billing give it flexibility that closed rivals lack. Running it well still needs serious GPU resources or cloud spend, and the polished consumer app experience is thinner than Kling or Runway. For developers and technical teams it is a compelling, differentiated option.

About Wan (Alibaba)

Wan is Alibaba's open-source AI video model family, with Wan 3.0 generating up to 30-second 1080p clips from text, images and documents, available open-source and via cloud API.

Wan is Alibaba's line of AI video models, developed alongside its Tongyi/Qwen ecosystem and released openly to the community. The latest Wan 3.0, launched August 24, 2026, generates videos up to 30 seconds, double its predecessor, at resolutions up to 1080p, and can build video from documents, spreadsheets, slides and web pages as well as text and images. Beyond generation, Wan carries forward editing abilities introduced in earlier versions, allowing changes to visuals, plot and dialogue, plus synchronized facial micro-expressions and multilingual voice output. Because core Wan models are open-source, developers can self-host and fine-tune them, while Alibaba Cloud's Model Studio offers managed API access billed per generated second. Wan is a genuinely different pick because it pairs a frontier commercial video model with an open-source release, making advanced text-to-video accessible to researchers, startups and self-hosters rather than only through a closed SaaS. It competes directly with Kling, Sora and Veo.

TL;DR

Wan is Alibaba's open-source AI video model family, with Wan 3.0 generating up to 30-second 1080p clips, available both open-source and via cloud API.

Company overview

Wan is developed by Alibaba as part of its Tongyi/Qwen AI program, released openly to build ecosystem adoption. It competes with Kling, Sora, Veo and other frontier video models.

Alibaba accelerated Wan investment in 2026 following a major share placement, launching Wan 3.0 in late August 2026.

Product features

Wan generates video from text, images and documents, produces up to 30-second 1080p clips, supports editing of visuals, plot and dialogue, and offers multilingual voice output.

Models are available open-source for self-hosting and through Alibaba Cloud Model Studio with per-second billing.

Target market

The market is AI developers, startups, researchers and technical video teams that want frontier video generation they can self-host or integrate via API.