Click to Skip Ad
Closing in...

Alibaba’s new Wan 2.1 text-to-video AI is unbelievable

Published Feb 26th, 2025 12:33PM EST
Alibaba's Wan website.
Image: Wan

If you buy through a BGR link, we may earn an affiliate commission, helping support our expert product labs.

One of the brilliant ideas that helped DeepSeek go viral was the Chinese company’s decision to make it open-source. That meant anyone could install DeepSeek AI on their computer for free without worrying about the official DeepSeek apps for iPhone or Android.

Well-known Chinese giant Alibaba is following in DeepSeek’s footsteps with a different type of AI software that also happens to be very exciting. Alibaba made its text-to-video AI service called “Wan” open-source on Wednesday. Wan competes directly against OpenAI’s Sora, an impressive AI video generation tool that was first unveiled last year and released a few months ago.

Specifically, we’re looking at the Wan 2.1 AI model, which will let users create videos with text, images, and even other videos in their prompts. Wan won’t just challenge OpenAI’s pricing model for Sora, but also OpenAI’s performance. Plus, Wan will be available for free because it is open-source. And perhaps most importantly, Wan 2.1 also tops the VBench leaderboard when it comes to performance. The videos it generates are so good, it’s hard to believe they were made by a free AI app.

The Wan2.1-T2V-14B AI model is currently ranked as the best-performing one. Alibaba has smaller models as well, which can be run on consumer hardware.

According to descriptions on the Wan website, the new AI service is capable of rendering “complex motion,” which involves creating “realistic videos featuring extensive body movements, complex rotations, dynamic scene transitions, and fluid camera motions.”

The website offers various examples of AI-generated videos, including a group of dogs riding bikes, two cats involved in a boxing match, and a team of dancers performing a choreography to prove this point.

Wan 2.1 can also generate videos that “accurately simulate real-world physics and realistic object interactions.” Alibaba offers additional genAI examples, including a woman splashing out of the water, an archer firing a bow, and a dog cutting tomatoes.

The text-to-video AI tool also supports “cinematic quality” videos, which means it should output “movie-like visuals with rich textures and a variety of stylized effects.”

Also impressive are the editing claims. Apparently, Wan supports precise edits using images and video references.

Finally, Wan 2.1 supports text generation in AI-generated videos. Alibaba says it is the first video model to support both Chinese and English text.

The Wan website also states that the program can generate sound effects and background music that match the visual content and the rhythm of the action.

In addition to the 14-billion model, Alibaba also released a Wan 2.1 T2V-1.3B model that needs only 8.19GB of VRAM to run. It will work with most consumer-grade GPUs and be fairly fast. “It can generate a 5-second 480P video on an RTX 4090 in about 4 minutes (without optimization techniques like quantization). Its performance is even comparable to some closed-source models,” the Wan website notes.

All of that sounds great, and the video samples available at this link are also quite incredible. Wan 2.1 definitely looks like like a frontier AI video generation tool that can compete against Sora and other similar rivals that come with a cost of access.

The fact that it’s open source means anyone can get started with Wan 2.1 as long as they know what they’re doing. Head on to Hugging Face and GitHub to start.

As you can see from the AI video examples I mentioned above, it’s easy to tell some of them are AI-generated. Others might fool viewers that they’re genuine clips. What I’m getting at is that tools like Sora and Wan can be used for nefarious purposes.

It’s great that Wan is open-source and others can inspect the code, but Alibaba’s website makes no mention of safety precautions. Also, it’s unclear how these sophisticated AI videos will be marked to inform users they’re watching AI-generated content.

Finally, I’ll remind you that Alibaba isn’t the only Chinese company developing an impressive AI video generation tool. A few days ago, ByteDance’s OmniHuman-1 AI impressed us with its capabilities.

Chris Smith Senior Writer

Chris Smith has been covering consumer electronics ever since the iPhone revolutionized the industry in 2007. When he’s not writing about the most recent tech news for BGR, he closely follows the events in Marvel’s Cinematic Universe and other blockbuster franchises.

Outside of work, you’ll catch him streaming new movies and TV shows, or training to run his next marathon.