tools
Wan 2.1 Review 2026 — Alibaba's Open-Source Video Model (Apache 2.0)
Wan 2.1 by Alibaba is a state-of-the-art open-source video model with Apache 2.0 license for commercial use. Text-to-video, image-to-video, 14B parameters.
Wan 2.1 provides local models and inference code for text-to-video, image-to-video, first-last-frame-to-video, video editing, text-to-image, and video-to-audio tasks. It includes 1.3B and 14B model variants, with 480P and 720P options depending on the task.
Researchers, developers, and creators can run it locally through Python, Gradio, ComfyUI, or Diffusers. The software and model weights are available under Apache 2.0; users provide their own compatible hardware. Optional remote prompt extension through DashScope requires its own API key.
Features
- Text-to-video generation with 1.3B and 14B models
- Image-to-video generation at 480P or 720P
- First-last-frame-to-video generation
- Video creation and editing through Wan2.1 VACE
- Text-to-image generation using the video model
- Video-to-audio generation
- Apache 2.0 license for the models and repository
- Local Gradio, ComfyUI, and Diffusers integrations
Use cases
- Generate short videos from written prompts
- Animate still images into video clips
- Create transitions between specified first and last frames
- Edit videos using text, masks, reference images, or conditioning inputs
- Run video-generation research on local GPUs
- Prototype image and video workflows with Python or Gradio