Models capable of generating visual content from text or images are gaining more and more relevance. One of the most recently presented is Wan 2.1, developed by Alibaba, one of the largest technology companies in China.
Launched in February 2025, this Chinese AI model has captured the attention of developers, creatives and companies for its ability to create realistic videos and images in an accessible and free way.
But what makes Wan 2.1 so special? Let’s explore in detail its features, applications, how to test it, and how it compares to other AI models.
What is Wan 2.1?
Wan 2.1 is an open source artificial intelligence model designed to generate videos and images from input text and images. Being open source means that anyone can use it, modify it and adapt it to their needs at no cost.
Its launch has been an important milestone in the democratization of AI, allowing both large companies and small developers to access advanced visual content generation tools.
Alibaba has made this model available on platforms such as Alibaba Cloud’s ModelScope and Hugging Face, facilitating its use globally. Additionally, its open source nature makes it an attractive option for those looking to customize and enhance the model based on their specific requirements.
##Wan Applications 2.1
The applications of Wan 2.1 are broad and can impact multiple industries. Below are some of the most notable ones:
- Marketing: Businesses can use Wan 2.1 to create personalized ads and promotional videos without the need for expensive production equipment. For example, it is possible to generate a video of a product in action simply by describing the scene in words, allowing for more dynamic and attractive campaigns.
- Entertainment: In the film and television industry, Wan 2.1 can revolutionize production by generating special effects, complete scenes or even short films, reducing both production times and costs.
- Education: This model is ideal for creating instructional videos or interactive simulations, such as science tutorials or historical reenactments, making learning more visual and accessible.
- Virtual Reality (VR): Wan 2.1 has great potential in creating content for VR environments, enhancing immersive experiences with AI-generated videos and images that can rival professional productions.
An unexpected aspect is its possible use in the creation of generative virtual worlds, an emerging trend in 2025 that includes immersive games and simulations. This opens up new possibilities for video game developers and metaverse platforms.
Wan 2.1 Technical Features
Wan 2.1 is built on advanced technology known as diffusion transformers, which allows large volumes of data to be processed to generate realistic visual results.
One of its key innovations is the use of a spatiotemporal variational autoencoder (VAE), which improves the model’s ability to handle complex movements and smooth transitions in the generated videos.
This means that Wan 2.1 can not only create static images, but also dynamic videos with interactions between multiple objects. For example, you can generate scenes like a dog riding a bicycle or a cat boxing, with a level of realism that impresses users.
##Wan 2.1 variants
The Wan 2.1 model is available in four different variants, each designed for specific tasks:
- T2V-1.3B: Generate videos from text with 1.3 billion parameters.
- T2V-14B: A more powerful version to generate videos from text, with 14 billion parameters, ideal for complex tasks.
- I2V-14B-720P: Generates videos from images with 720P resolution.
- I2V-14B-480P: Similar to the previous one, but with 480P resolution, offering a balance between quality and efficiency.
T2V variants focus on generating videos from textual descriptions, while I2V allow creating videos from static images.
How to Test Wan 2.1?
The easiest and most direct way to try Wan 2.1 is by visiting its official platform: wan.video.
This site allows you to experiment with Alibaba’s AI model to create videos and images from input text or images, and is designed to be accessible regardless of your level of technical experience. To get started, follow these steps:
- Go to wan.video.
- Register or log in if the platform requires it.
- Follow the instructions that appear on the site to generate visual content with Wan 2.1.
It is an intuitive, free and open source option, ideal for both beginners and advanced users who want to explore its possibilities at no cost.
Secondarily, you can also try Wan 2.1 on platforms like Alibaba Cloud’s ModelScope and Hugging Face, where the model is available for use or customization.
##How does it compare to other AI models?
It is common to compare Wan 2.1 with other popular AI models, such as ChatGPT or Gemini. However, it is important to note that these are language models, focused on generating and understanding text, not creating visual content. Therefore, Wan 2.1 does not directly compete with them, but rather offers complementary functionality.
A more comparable model is OpenAI’s Sora, which also generates videos from text. According to benchmarks like VBench, Wan 2.1 scores about 84.7% overall, outperforming Sora in key metrics like scene generation quality and accuracy on individual objects.
Additionally, while Sora is a proprietary model with restricted access, Wan 2.1 is open source, making it more accessible and customizable for a wide range of users.
Other Alibaba Products: Qwen and EMO
Alibaba has not limited itself to developing Wan 2.1; It has also created other AI models that complement its technological ecosystem:
Qwen
It is a family of large language models (LLMs) that includes variants such as Qwen 2.5, released in January 2025. This model has outperformed competitors in reasoning, language understanding and code generation tasks, establishing itself as a powerful tool for developers and companies.
EMO
Released in February 2024, EMO is a framework for generating talking head videos from a single image and an audio file. This allows the creation of realistic digital avatars that can speak or sing, with applications in entertainment, social media and personalized content creation.
Without a doubt, this is just the beginning of what Alibaba and other tech giants have in store for the future of AI.
Whether you’re looking to create dynamic ads, produce educational content, or explore new forms of entertainment, Wan 2.1 offers a unique opportunity to innovate without the traditional barriers of cost and complexity.
This post is also available in: