Alibaba open-sources AI model for video and image generation

Alibaba Group Holding Ltd. has announced that its video and image-generating AI model, Wan 2.1, is now publicly available as an open-source platform. The move is expected to boost adoption and heighten competition in the AI space.

Alibaba’s decision aligns with a similar strategy by ByteDance and startup DeepSeek. Earlier this year, DeepSeek’s affordable open-source models garnered attention from tech investors and surprised the capital-intensive industry with performance comparable to established players like OpenAI.

Alibaba expands AI offerings with four Wan 2.1 variants

Alibaba has released four variants of its Wan 2.1 AI model—T2V-1.3B, T2V-14B, I2V-14B-720P, and I2V-14B-480P—which generate images and videos from text and image inputs. The “14B” designation signifies that these variants process 14 billion parameters, enabling them to handle more complex inputs and produce more accurate results.

The models are available globally on Alibaba Cloud’s ModelScope and HuggingFace platforms for academic, research and commercial use.

Alibaba introduced the latest version of its video and image generating AI model in January, later shortening its name to Wan from Wanx. The company touted its ability to generate highly realistic visuals. Alibaba has since emphasized its top ranking on VBench, a leaderboard for video-generative models, where it excels in functionalities like multi-object interaction.

On Tuesday, the company unveiled a preview of its reasoning model, QwQ-Max, which it intends to make open-source upon full release.

ByteDance launches OmniHuman, an AI that creates full-body videos from one image

The surge in open-source AI development comes amid significant advancements from competitors. ByteDance, the parent company of TikTok, recently announced the launch of OmniHuman , an advanced AI system that can generate an entire full-body video from the input of just one photograph.

What sets OmniHuman apart from similar AI systems is that it can create dynamic videos in which the subject can speak, gesture, and move naturally. The system has been trained on over 19,000 hours of video content and uses this data to create videos that mimic human behavior in terms of movements and expressions.

One of the first demonstrations using OmniHuman is a video of Albert Einstein delivering a lecture. The video went viral the moment it was shared on the Internet, which is proof of how much of a difference this feature makes.

While OmniHuman is the first AI system to claim to generate a believable video with just one image, other systems can deliver similar results using a few images.

For starters, there’s Pollo AI, which lets users generate videos by uploading images. It is easy to use, and with just images and a text prompt, it can create videos based on user-defined styles.

Canva has a similar AI Video Generator. Its set of templates and editing features make it easy for users without advanced video creation skills to create AI-generated videos using images.

Freepik also has a similar AI Video Generator that offers several generation modes and customization options to create engaging videos with just images and prompts.

Alibaba commits $52 Billion to AI and cloud expansion

A few days ago, Alibaba announced plans to invest at least 380 billion yuan ($52 billion) over the next three years to strengthen its cloud computing and AI infrastructure.

The internet giant, whose co-founder is Jack Ma, will invest more in its artificial intelligence and cloud computing infrastructure than it has for the past 10 years. With evolving models becoming more compute-intensive, the company said on its official blog that Alibaba plans to stand out as a partner to companies building and deploying AI in the real world.

This aim is one of China’s largest investments in AI infrastructure, and it underscores Alibaba’s growing ambitions in the field. But it comes amid investor worries over whether some big tech firms will overestimate demand for AI services in the future or underestimate the capital cost of building them.

TD Cowen analysts pointed out Friday that Microsoft Corp. is cancelling leases for substantial data center capacity in the US. This move may reflect concerns about whether it’s building more AI computing than it will need over the long term. Alibaba’s Hong Kong shares slid as much as 3% on Monday.

Other Big tech firms, such as Meta Platforms Inc. and Amazon.com Inc., are pledging billions to build the data centers needed to train, develop, and host AI services.

Cryptopolitan Academy: Want to grow your money in 2025? Learn how to do it with DeFi in our upcoming webclass. Save Your Spot