In September 2024, Alibaba officially released its proprietary AI video generation model, Tongyi Wanxiang AI, also referred to as Wanx AI. Come January 2025, Alibaba introduced Wanx 2.1, the latest iteration of its AI video generator.
Now, they face stiff competition from tech companies like OpenAI and Kuaishou, but does Alibaba have what it takes to stand out?
To answer this, I compiled an in-depth guide to provide insight into what makes Wanx AI such a big deal in AI visual content creation.
Wanx 2.1: The Basics
Wanx 2.1 uses a mix of VAE and DiT technology to make videos look super realistic by improving how things move and connect visually. Basically, it’s great at copying tricky real-world movements with spot-on body coordination and smooth motion.
This means that I can use it to render complex character scenes like ballerina dancing, swimming, and figure skating, which most AI video models often fail at. In fact, it is because of its capacity to adhere to realistic motion trajectories that Wan 2.1 stands at the top of the AI video generation VBench leaderboard.
Aside from that, this new version comes with even greater prompting capabilities, which lead to faster and more intuitive generations. For this reason, I can generate 1-minute videos in 1080p resolution within 15 seconds or so. It’s also worth pointing out that Wan 2.1 has four variants: T2V-1.3B, T2V-14B, I2V-14B-480P, and I2V-14B-720P.
Since it supports 14 billion parameters (14B), the AI video model can interpret far more input and context than before. In February 2025, it was announced that all four variants are now open-source. This makes Wan 2.1 one of the few AI video models that can be freely accessed and modified by public users and developers.
What Is My Personal Opinion of Wan 2.1?
I gave Wan 2.1 a shot by testing it with a few sample videos, and I have some mixed feelings about it. For my first try, my first prompt was: "Two massive dragons engage in an epic aerial battle over a medieval kingdom, unleashing fire and chaos, with the camera panning to show the destruction below."
Here is the generated video:
The scene looked great—destruction and all—but the dragons? Not so much. They just hovered face-to-face in the sky, doing nothing, which made the motion feel stiff and disappointing.
I tried again with a more detailed prompt: "Two massive dragons clash over a medieval kingdom, scales glinting as one dives with claws slashing and the other counters with a fiery blast, wings beating as they spiral and dodge through smoky skies, tails whipping with realistic force, while the camera shifts smoothly between wide shots of the kingdom and close-ups of the fight."
This time, the video was way better—the dragons’ movements were dynamic and intense, with natural physics, and the camera transitions felt smooth and lively.
In my opinion, Wan 2.1 has potential, especially since it uses VAE and DiT tech to handle realistic motion well. But it really needs detailed prompts to deliver; otherwise, the motion can feel flat, which was a bit annoying at first. With some effort, though, it can create awesome, dynamic videos.
What Features Do I Like Most About Wan 2.1?
I can’t deny that Wan 2.1 introduces a wide range of advancements that take Alibaba’s AI video solution to the next level, even when stacked up against other AI video tools. So, let me break down the AI model’s key strengths that make it such a standout in my view:
Superior Performance
Wan 2.1 employs proprietary VAE technology that enables it to reconstruct high-resolution 1080p videos without compromising on smooth motion. As I mentioned before, it also preserves visual detail well, so the frame-to-frame coherence is relatively good.
In other words, there’s less risk of having to worry about flickering or distortion across frames. On top of that, Wan 2.1’s VAE architecture can encode and decode videos at an incredibly fast rate. This means I can rely on it to help with near-real-time video creation.
Multilingual Understanding
Wan 2.1 is the first AI video generation model capable of understanding text prompts in both English and Chinese native languages. This bilingual feature can be fantastic for producing animated texts and all sorts of overlays in the videos.
I can also use Wan 2.1 to potentially craft prompts for product videos or even interactive tutorials for native audiences with much more effective results. Plus, these robust text generation capabilities give it a fair advantage over other AI video models.
Unmatched Motion Dynamics
Wan 2.1 has an impressive mastery over motion dynamics in AI video generation. While I don’t think it necessarily leads on visual aesthetics, this AI video model maintains an undeniable balance between scene consistency, motion realism, and spatial precision.
For the most part, this makes Wan 2.1 well-suited for generating professional-grade visuals that look and feel realistic. Whether it’s trailers, music videos, animated scenes, or even gaming assets, I’m confident it can deliver smooth and believable results.
Open-Source Accessibility
Alibaba chose to release Wan 2.1 as a free and open-source solution, which includes all four variants. I really appreciate this because it effectively makes it more accessible to businesses, brands, developers, and creators worldwide.
It becomes easy to integrate Wan 2.1 and automate all sorts of complex video creation tasks, even if you lack any coding expertise. Plus, I like how the lower barrier to entry means that it will help foster innovation in the wider AI community.
How Do I Prefer To Access Wan 2.1? Introducing Pollo AI
You can access Wan 2.1 by installing it locally or via the developer's official website, Wan.Video. However, I should tell you that these aren’t the easiest ways to use the AI video model.
Instead, I would suggest you consider using Pollo AI. This is a cutting-edge, all-in-one AI image and video generation platform, integrated with several industry-leading AI models. Some of these include Runway, Kling AI, Pixverse, Hailuo, Luma AI, and, of course, Wanx AI,
Since they are all in one place, this makes it easy to directly compare the video outputs between models.
Besides that, I can access numerous AI tools and templates on Pollo AI that make it easy to create all kinds of custom videos in a flash.
Best of all, the platform has very affordable pricing plans, so I didn’t have to break the bank to enjoy all its unique features and tools. But you don’t have to take my word for it! Check out Pollo AI at no cost via its free trial now!
My Final Say on Wan 2.1
I find that Wan 2.1 can help any creator produce realistic and believable character videos in almost any visual style. It still faces stiff competition against other rivals like Kling AI, but it remains undefeated in terms of dynamic motion and pattern consistency across scenes. Head over to Pollo AI now and start generating videos with Wan 2.1 to see what it can do for you!