Wan

Wan is an open-source AI video generation platform developed by Alibaba, offering text-to-image, image-to-image, text-to-video, and image-to-video creation through advanced multimodal technology. Open-sourced under the Apache 2.0 license, it aims to lower the barrier to AI creation so ordinary users can run high-quality video generation models on consumer GPUs.

Key Features

• Multimodal content generation: Supports text-to-image, image-to-image, text-to-video, image-to-video, and other creation modes
• High-quality video output: The latest Wan 2.5 version supports generating video at 480p to 1080p resolution with synchronized audio (dialogue, sound effects, and background sound)
• Open source and deployable: Offers a 14B professional edition and a 1.3B lightweight edition - the lightweight version needs only 8GB of VRAM to run on consumer cards like the RTX 4090
• ComfyUI integration: Native ComfyUI workflow support with one-click deployment packages and full tutorials for a quick start
• Multilingual text rendering: The first AI video model to support rendering visible Chinese and English text in video

Use Cases In content creation, video makers can quickly generate short video material from simple text or image prompts, dramatically shortening creative cycles; in social media operations, creators can use image-to-video to turn static assets into dynamic content; and academic researchers and developers can optimize and customize the models based on the open-source code.

Unique Advantages Wan's biggest strengths are its open-source nature and hardware friendliness. Compared with closed-source commercial models like Google Veo 3, Wan 2.1 scored a high 86.22% on the VBench benchmark, outperforming several well-known models including Sora and Hunyuan while keeping the running requirements extremely low. Wan 2.5 goes further as an early leader in synchronized video-and-audio generation, delivering a near-commercial-grade creation experience at an affordable price.