Wan 2.2 (ComfyUI) review

What is Wan 2.2?

Wan 2.2 is a family of open-weight video models from Alibaba’s Wan team. The weights are free under the Apache 2.0 license, which also allows commercial use. You run them on your own computer, usually through ComfyUI, a node-based tool that has a guide for Wan 2.2. There is no service in between, so there is no account, no age gate, no per-clip fee and no content filter beyond the license terms. In return, you supply the hardware, the setup time and the judgment. The 14B models need far more video memory than the 5B model.

Key features

Several input modes

You can start from text, an image, text plus an image, speech audio with a reference image (S2V), or a driving video with the Animate model. Few hosted tools accept all of these.

Your own length and resolution

You set the frame count in ComfyUI, and no length cap is published. S2V clips follow the length of the audio you give them. You get 480p or 720p output; the TI2V-5B model renders 720p at 24 fps.

Character swap

Image-to-video keeps the look of a fixed still. The Animate model has a replacement mode that swaps a character into an existing video.

Full privacy

After the local install, nothing needs to leave your machine. You can run it offline.

The catch

You need a strong GPU. Natively the 5B model wants 24 GB of VRAM; ComfyUI offloading cuts that to about 8 GB, and the 14B models need much more. Setup is technical. Wan 2.2 does not generate audio; S2V only uses audio you supply. With no service rules, the license still bans illegal or harmful content, and you carry full legal responsibility for what you make.

Final verdict

Wan 2.2 suits you if you own a capable GPU and are comfortable with ComfyUI. You get no per-clip cost, complete privacy and a wide set of input modes. If you do not have the hardware, the free license does not make it cheap, and the missing audio means extra work in another tool. For everyone else, a hosted service is the faster path.