Wan 2.2

Alibaba's Apache 2.0 video models: a two-expert 27B text and image to video family, plus a 5B model for a 24 GB GPU.

In short. Wan 2.2 is a set of AI video generators from Alibaba that anyone can download. Give it a text prompt or a picture and it makes a short clip, five seconds at up to 720p. The big models need a data-center GPU, but the 5B model can run on a single gaming GPU with 24 GB. It uses the Apache 2.0 license, so commercial use is allowed. It is the newest general-purpose Wan you can download.

Summary

Alibaba released the Wan 2.2 code and weights on 28 July 2025 under the Apache 2.0 license. The family has three models: T2V-A14B for text-to-video, I2V-A14B for image-to-video, and TI2V-5B, a single model that does both.

The two A14B models use a mixture of experts designed for the denoising process of a video diffusion model. A high-noise expert handles the early steps that set the overall layout, and a low-noise expert handles the later steps that refine detail. Each expert has about 14 billion parameters, so the model has about 27 billion in total but only 14 billion are active at each step, which keeps compute close to a single 14B model. They produce clips of about five seconds at 480p or 720p. The 5B model uses a high-compression video autoencoder (16 by 16 by 4) and generates 720p video at 24 frames per second.

The weights are shipped in full 32-bit precision: about 126 GB for each A14B model, including the UMT5 text encoder, and about 34 GB for the 5B model. Alibaba's single-GPU instructions call for at least 80 GB of VRAM for A14B and at least 24 GB, such as an RTX 4090, for the 5B model with offloading enabled. Community toolkits such as LightX2V and DiffSynth-Studio add quantized versions and low-memory offloading for smaller cards. Wan 2.2 works with ComfyUI and Diffusers.

Alibaba later released more models on the Wan 2.2 base: Wan2.2-S2V-14B for audio-driven video (August 2025), Wan2.2-Animate-14B for character animation and replacement (September 2025), Wan-Dancer-14B for music-to-dance (July 2026) and Wan-Animate-2 (August 2026), all under Apache 2.0. The newer general-purpose versions, Wan 2.5, 2.6, 2.7 and 3.0, are not open: as of 8 November 2026 the Wan-AI page on Hugging Face lists no weights for them, and they are offered through Alibaba's apps and API. Alibaba's claim that Wan 2.2 beats leading commercial models comes from its own Wan-Bench 2.0 and has not been independently verified.

Good for

  • Local, commercial-friendly video generation under Apache 2.0, with no revenue limit
  • Image-to-video and text-to-video experiments on a 24 GB GPU using the 5B model
  • Research, fine-tuning and LoRA work, since there is a large community ecosystem
  • A base for character animation, speech-to-video and dance models from the same team

Watch out for

  • The A14B models need an 80 GB GPU by Alibaba's instructions, or quantized community builds and offloading
  • Clips are short, about five seconds, and the base models make video without sound
  • It is more than a year old, and newer open models such as LTX-2.5 add synchronized audio and multi-shot scenes
  • Wan 2.5 and later are API-only, so do not trust download links for them
  • Alibaba's quality claims rest on its own benchmark

Performance claims come from the developer and have not been independently verified.

How much memory does it need?

Memory figures for each variant of Wan 2.2. Image, video, speech and music pipelines also load a text encoder and a decoder, so the memory needed can be larger than the main checkpoint.

Variants Parameters Download size Memory needed Fits on
TI2V-5B (text or image to video) FP32 5B 34 GB 24 GB Gaming GPU
T2V-A14B (text to video) FP32 27B 14B active 126 GB 80 GB Workstation
I2V-A14B (image to video) FP32 27B 14B active 126 GB 80 GB Workstation

Download size covers every component needed to run the model, such as the generator, text encoder and decoder. Memory needed is a stated or published measured minimum where one exists; otherwise it is the download size plus about 15%. Larger images, longer clips and higher resolutions need more. The developer states a minimum of 24 GB of GPU memory.

The license in plain words

One of the most widely used permissive licenses. You may use, change and sell products built on the model with only light conditions.

More about Apache 2.0

See also

Frequently asked questions

What is Wan 2.2?
Wan 2.2 is a set of AI video generators from Alibaba that anyone can download. Give it a text prompt or a picture and it makes a short clip, five seconds at up to 720p. The big models need a data-center GPU, but the 5B model can run on a single gaming GPU with 24 GB. It uses the Apache 2.0 license, so commercial use is allowed. It is the newest general-purpose Wan you can download.
Can I use Wan 2.2 commercially?
Yes. Wan 2.2 is released under Apache 2.0, a permissive license. You can use it commercially if you follow the standard conditions, such as keeping the license notice.
Is Wan 2.2 open source?
Wan 2.2 is released under Apache 2.0, a license approved by the Open Source Initiative. Training data and training code may not be public, so it is best described as open weights under an open-source license.
Can I run Wan 2.2 on my own computer?
Yes, on a good graphics card. The smallest variant, TI2V-5B (text or image to video), needs about 24 GB of memory at 4-bit precision, which suits a 24 to 32 GB gaming GPU or a Mac with 32 GB or more.
When was Wan 2.2 released?
Wan 2.2 was released by Alibaba Wan on July 28, 2025.

Official links

Sources

Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.

Release date and facts last verified November 8, 2026.

Related models

Model Released Parameters Context License Smallest variant needs
LTX-2.5 LTX (Lightricks) 22B – LTX-2.x Community (Conditional) 16 GB Laptop or phone
FLUX 3 Image Klein Black Forest Labs 5.6B – FLUX Kommunity (Conditional) 8 GB Laptop or phone
MiniMax H3 MiniMax 33.1B – MiniMax H3 Community (Conditional) 166 GB Big-memory desktop
Pika H3 Flex Pika 33.1B – MiniMax H3 Community (Conditional) 169 GB Big-memory desktop