Wan2.2
An open source AI video model family generates clips locally on GPUs from text, images or speech and supports character animation under Apache 2.0.
Wan2.2 is a family of local AI video models for creators and developers who want to generate clips on their own hardware. Different models handle text, still images, speech and character animation.
- Text and image inputs: Generate clips from a written description or a still image, with dedicated models for each task and TI2V-5B for both.
- Speech-driven video: S2V-14B generates video from speech and supports output up to 720p.
- Character animation: Animate-14B supports animating characters and replacing them in video.
- Visual style: The models support control over the visual style of generated clips, with training data labeled for lighting, composition, contrast and color tone.
The open source project uses the Apache 2.0 license, and model weights are available through Hugging Face and ModelScope. TI2V-5B can run on a single consumer GPU and produce 720p video at 24 frames per second. The larger models also support execution across multiple GPUs.
Logotyper
