What Is Wan AI? Get to Know Alibaba’s AI Video Generator

The development of AI video generators has made the video production process increasingly easier. One model that has received significant attention is Wan AI, a family of generative video models from Alibaba developed as an open model for various video creation needs.

One of its important versions is Wan 2.2, which brings a Mixture-of-Experts (MoE) architecture, improved training data, and the ability to generate videos up to 720P on certain models. Wan 2.2 also supports video generation from both text and images.

Key Takeaways

• Wan AI is a family of generative video models developed by Alibaba, with Wan 2.1 and Wan 2.2 as its main open-source versions.
• Wan 2.2 uses a MoE architecture and supports text-to-video and image-to-video up to 720P on certain models.
• Hardware requirements depend on the model. Wan 2.1 has a T2V-1.3B variant requiring around 8.19 GB of VRAM, while Wan 2.2 has larger models.

What Is Wan AI?

Wan AI is a family of generative models for creating and processing videos using artificial intelligence. 

The model was developed by the Alibaba team and is available as an open model that can be used by developers and researchers.

Wan 2.1 became one of the important releases because it offers several capabilities at once, including text-to-video, image-to-video, video editing, text-to-image, and video-to-audio. The model was also designed so that some of its variants can run on consumer GPUs.

Meanwhile, Wan 2.2 is the next development, focusing on visual quality, more complex motion, and cinematic control. Alibaba released Wan 2.2 in July 2025 as an open-source generative video model based on MoE.

Read also: AI Image Generators 2026: Comparison of GPT Image 1.5, Flux 2 Max & Gemini 3 Pro

What Is Wan 2.2?

Wan 2.2 is the next generation of the Wan video model. One of its main technical changes is the use of Mixture-of-Experts (MoE). 

This approach uses different expert models at certain denoising stages, allowing model capacity to be increased without proportionally increasing the computational workload.

Wan 2.2 was also trained using a larger dataset compared with Wan 2.1. Project documentation states that the number of images in the training data increased by 65.6% and videos by 83.2%. 

This data is used to improve the model's ability to understand motion, semantics, and visual aesthetics.

One interesting model is Wan2.2-TI2V-5B, which combines text-to-video and image-to-video in a single model. This model supports 720P video at 24 fps and can run on consumer GPUs such as the RTX 4090 with the appropriate configuration.

Be productive with AI, invest smartly. Register on Bittime and build your portfolio now.

Key Features of Wan 2.2

Wan 2.2 does not focus only on text-to-video. Its model family covers several video creation needs.

Text-to-video allows users to provide a prompt to generate a video scene. Image-to-video uses an image as the starting point and then generates motion based on instructions.

In addition, models are available for needs such as speech-to-video and character animation. The Wan 2.2 model documentation lists T2V-A14B, I2V-A14B, TI2V-5B, S2V-14B, and Animate-14B, each with different functions.

Wan 2.2 also provides more detailed visual control. Alibaba states that the model was trained using data with aesthetic labels such as lighting, composition, color, and camera characteristics. 

This helps users provide more specific visual instructions through prompts.

Read Also: What Is Lucy Decart AI? AI Video Technology for Real-Time Virtual Worlds

Is Wan AI Free?

The question of whether Wan 2.2 is free needs to distinguish between the model and online services.

Wan 2.2 is an open-source/open-weight model that can be used locally with suitable hardware and configurations. 

In this scenario, users do not pay a per-generation fee to a service provider, but they still need their own GPU, storage, and computing resources.

Meanwhile, online platforms that provide access to Wan AI may implement credit systems, usage limits, or paid plans. Some services do offer free access, but their terms and capacity vary.

Alibaba Cloud itself provides Wan services through Model Studio with different models and usage schemes. Therefore, users should check the platform they are using before assuming that all Wan AI services are free or unlimited.

Don't miss updates on AI coin prices such as Bittensor (TAO), Venice Token (VVV), NEAR Protocol (NEAR), and Internet Computer (ICP) on Bittime.

How to Use Wan AI to Create Videos

How to use Wan AI depends on the platform selected. In general, the process consists of several stages.

First, determine whether the video will be created from text or an image. For text-to-video, write a prompt describing the subject, environment, motion, lighting, and camera angle.

Second, choose the appropriate model and resolution. Wan 2.2 has several models with different capabilities, so hardware requirements also vary.

Third, run the generation process and review the result. More specific prompts can help direct camera movement, characters, and the visual atmosphere.

For users without a GPU with sufficient specifications, cloud services or online platforms can be an alternative because the computing process is handled on servers.

Read Also: Meet Muse AI, Meta's App That Surpassed ChatGPT on the U.S. App Store

Who Is Wan AI Suitable For?

Wan AI can be used for various visual needs, from creating scene concepts, storyboards, and social media content to creative experiments and AI development research.

For developers, open models provide the flexibility to run models locally and integrate them into their own workflows. 

For creators, meanwhile, text-to-video and image-to-video capabilities can be used to create short visuals without having to start the entire process with a physical camera.

However, generation results still depend on the model, prompt, hardware, and configuration used. AI video can also produce motion or details that do not always match the instructions, so review and editing are still necessary.

Conclusion

Wan AI is a family of generative video models from Alibaba that offers text-to-video, image-to-video, and various other video workflows. 

Wan 2.2 brings improvements through its MoE architecture, larger training data, and more detailed visual controls.

An important advantage of Wan is the availability of open models and support for consumer hardware on certain models. However, GPU requirements differ between Wan 2.1, Wan 2.2, and other model variants.

For users who want to try an AI video generator, Wan 2.2 can be one option to explore, especially if flexibility between text-to-video and image-to-video is needed.

Interested in investing in the AI crypto sector? Monitor prices and start trading popular AI coins such as Bittensor (TAO), NEAR Protocol (NEAR), and Artificial Superintelligence Alliance (FET) directly on Bittime

Bittime is a licensed and OJK-supervised Digital Financial Asset Trader (PAKD) platform — where you can buy Bitcoin in Indonesia and hundreds of other crypto assets starting from Rp10,000. The registration process is fast, secure, and can be started today.

Track the USDT to IDR conversion and the real-time price movements of your favorite crypto assets. Everything is available in one crypto investment app that can be downloaded for free from the Play Store.

Ready to get started? Register on Bittime now and execute your investment strategy with a platform trusted by millions of users in Indonesia.

FAQ

What Is Wan AI?

Wan AI is a family of generative video models from Alibaba that can be used to create and process videos using text, images, and other formats as inputs.

What Is Wan 2.2?

Wan 2.2 is a generation of the Wan video model that uses a Mixture-of-Experts architecture and supports various tasks such as text-to-video and image-to-video.

Is Wan 2.2 Free?

The Wan 2.2 model is available as an open model that can be run locally. However, online services that provide access to Wan may have different fees, credits, or usage limits.

Can Wan 2.2 Create Videos from Images?

Yes. Wan 2.2 has image-to-video and TI2V models that allow images to be used as inputs to generate videos.

Does Wan 2.2 Require a GPU?

If run locally, Wan 2.2 requires a GPU with sufficient capacity for the model and configuration being used. One TI2V-5B model supports 720P and can run on consumer GPUs such as the RTX 4090 with the appropriate configuration.

What Is the Difference Between Wan 2.1 and Wan 2.2?

Wan 2.2 introduces an MoE architecture, a larger training dataset, and improved aesthetic control and video capabilities compared with Wan 2.1.